# Croissant & Baguette > Websites and stores that load in under a second. Apps built around how you actually work. AI systems that handle the enquiries, bookings and paperwork your team shouldn't be doing. Built in Sydney since 2019, and yours to keep. Sydney, Australia. https://croissantbaguette.com.au ## Croissant & Baguette https://croissantbaguette.com.au/ Sydney studio building AI systems that run your bookings, enquiries and back-office, on infrastructure you own. Plus growth and web builds. Since 2019. ### We build what your business runs on. Websites and stores that load in under a second. Apps built around how you actually work. AI systems that handle the enquiries, bookings and paperwork your team shouldn't be doing. Built in Sydney since 2019, and yours to keep. - **38× faster** — 27s → 0.7s - **6 days from kick-off to live store** — after a week of design, January 2026 - **+37% conversion rate** — and +15% cross-sell, the second tour sold in the cart - **3 → 1 sites, one codebase** — three brands from one repository and one CMS - **5× faster** — fully loaded in 0.9s, from 4.6s - **Double digit growth in sales** — since the move off the theme, by the client's count - **<0.5s page loads** — perfect performance scores on a brand-new domain - **3× faster** — fully loaded in 549ms, from 1.6s - **10k SKUs, every page under 1s** — category pages took up to 20 seconds on the old build - **33% conversion rate** — Shopify, August 2026 - **24/7 leads qualified by AI** — every social message and website enquiry read, sorted and logged in HubSpot - **0.8s fully loaded** — down from 10 seconds or more on the old WordPress site - **30 days from kick-off to an Android app on live phones** — app, venue CMS and fleet cockpit in 22 staged releases ### Pick your entrance. - **AI Automation** Work that shouldn't be done by hand, handled end-to-end. Enquiries answered and booked. Documents checked and filed. Systems that finally talk to each other. Built to run, not to demo. - **SEO & AI Visibility** Customers now ask ChatGPT, Claude and Google's AI before they call anyone. We measure whether you're the answer, fix what stops you, and report the number every month. - **Web Development** Websites, stores and web apps that load in under a second and don't need a developer to update. The delivery layer for everything above. ### Most businesses use AI. Almost nothing runs on it. ChatGPT writes the email; someone still copies it into the CRM, checks the calendar and files the form. We build the part that runs, connected to the tools you already have, with a human approving what matters. ### Everything we make. - **Web Development** - **A website that loads in under a second.** Pre-rendered or WordPress, designed from scratch, edited by your team without calling us. - **A store that's actually yours.** Shopify under the hood, with a headless storefront or a custom theme on top. - **An app built around how you work.** Portals, bookings, dashboards, internal tools, and the mobile app your team carries. - **SEO & AI Visibility** - **Being the answer.** Customers ask ChatGPT, Claude or Google's AI before they call anyone. We measure whether they hear your name, and fix what stops them. - **A number every month.** How often you're cited across the engines, against your competitors, and what we did to move it. - **AI Automation** - **Enquiries & bookings.** Answered, qualified and booked into your calendar and CRM. Humans handle the exceptions. - **Documents & applications.** Read, checked against your rules, extracted and filed. Flagged when something's off. - **Systems that don't talk.** Booking tool, accounting, CRM and spreadsheets kept in sync. No re-keying. ### You keep everything. The code, the data, the final say. We build it; it belongs to you. - **You own the code.** No per-task fees, no platform we can hold you to. - **Your data stays in Australia.** Self-hosted or Australian-region cloud. Your call. - **A human signs off.** Anything that matters waits for approval. Every action is logged. - **Built by the people you talk to.** Three of us in Sydney, shipping production software since 2019. ### Recently served. - Booking platform on Rezdy, 10s to under 2s, conversion rate up 37%. (https://croissantbaguette.com.au/projects/cheeky-kiwi-travel/) - ASX-listed financial services, 340 pages, published by their own team. (https://croissantbaguette.com.au/projects/fiducian/) - Headless Shopify store in six days. First sale in week one, 33% conversion rate. (https://croissantbaguette.com.au/projects/made-quiet/) - Free scanner to paid launch engine, extension and API, in nine weeks. (https://croissantbaguette.com.au/projects/nilkick/) - 27s to 0.7s. Same content, new build. (https://croissantbaguette.com.au/projects/wingboy/) - Headless Shopify store and the portal that pays 343 makers. (https://croissantbaguette.com.au/projects/pop/) - "The new website received positive feedback from customers: they commended the site's speed, fluidity, and overall UX. Croissant & Baguette worked closely with the client, providing regular updates and constant support throughout. Also, their knowledge and technical expertise were impressive." — Arani Satgunaseelan, CEO, Vollardian - "The client was pleased with the outcome of the engagement. Croissant & Baguette managed the project well: everything was delivered on time, and all queries and concerns were addressed promptly. They also impressed the internal team with their transparency and strong adherence to quality." — Rejimon Punchayil, CEO, Karis Life - "Croissant & Baguette successfully designed and deployed the platform. Croissant & Baguette led an excellent process with on-time deliveries and a responsive approach. The client also praised the team's ability to understand their vision for the product." — Natalie, CEO, TTMF ### Tell us what you need. A site, a store, an app, or the admin your team shouldn't be doing. Two lines is enough. You'll get a straight answer and a next step within a business day. ## AI Automation https://croissantbaguette.com.au/ai-automation/ Sydney AI automation: systems that run enquiries, documents and bookings end-to-end, on infrastructure you own, with a human in the loop. From $8k. **Service:** AI automation (Business process automation) We build AI systems that take over a defined part of a business end-to-end: the enquiries, the documents, the handoffs between systems. Delivered on infrastructure the client owns, with a human approving anything that matters, and kept running under a managed plan. - **Automation Opportunity Assessment** from A$1500 — Two weeks mapping the manual work in your business, ranked by hours and risk, with build prices attached. $1,500 fixed, credited against a build started within 90 days. - **Automation build** from A$8000 — One process automated end-to-end, integrated with what you already run, with 30 days of hypercare. From $8,000, typically $10–30k per process. - **Managed** from A$250 — Monitoring, fixes when models or APIs change, and a monthly report of what the system handled, saved and got wrong. From $250 a month. - **Internal tool / dashboard** from A$8000 — Operations dashboards on top of the systems doing the work. Typically $8–20k. ### Work that shouldn't be done by hand. We build AI systems that take over a defined part of your business end-to-end: the enquiries, the documents, the handoffs between your systems. Then we keep them running. You own the code, your data stays in Australia, and a human approves anything that matters. ### Using AI isn't the same as running on it. Most Australian businesses now use ChatGPT for emails and drafts. Almost none have changed how an enquiry gets answered, how a document becomes a filed record, or how two systems stay in sync. The admin is still admin; it's just done with better prose. The gap between "we use AI" and "this runs itself" is where the hours are, and closing it is an engineering job, not a prompt. ### Five jobs we automate. Each one is a pattern we've shipped, shown with the system that runs it. If your version looks different, that's exactly what the assessment is for. - **Enquiries & bookings** An enquiry lands at 9:40pm. The system reads it, answers in plain English, checks real availability, holds the spot, writes the booking into the calendar and the CRM, and sends the confirmation. Anything unusual waits in a queue for a person, with a drafted reply attached. — For tour operators, clinics, education providers, and anyone else whose inbox is a to-do list. — A booking platform on the Rezdy API with an AI admin behind it that drafts the daily supplier emails from live bookings and answers booking questions in plain English, from the live data only. — Cheeky Kiwi Travel — The 3pm supplier run took Cheeky Kiwi staff one to two hours a day for fifteen-plus suppliers. Now Claude drafts every email from tomorrow's bookings and a person reviews and sends. - **Documents & reports** A client report that someone would assemble by hand gets built by the system instead: it pulls the data from the source, writes the document, finds the client's record, and either sends it or leaves it for one person to approve. The same thing works in reverse. What arrives gets read, checked against your rules, extracted into the file, and the missing pieces chased. — For brokers, advisers, agents, legal, licensing, and any service that runs on paperwork. — "Send Sarah the report on 12 Smith Street" is one sentence: the report is generated on the property data service, matched to Sarah's record in HubSpot, built and emailed. — Joris — Joris asks in the Claude and ChatGPT he already pays for, and our MCP server does the work behind it. Nothing new for him to log into, and switching on a new ability doesn't mean an update he has to install. - **Systems that don't talk** A booking changes in one place. The invoice updates in Xero, the CRM record follows, the ops spreadsheet stops being wrong, and nobody re-keys anything. "Which one is right?" stops being a question anyone asks. — For businesses running a booking tool + accounting + CRM + spreadsheets. — Every sale, in store or online, flows from Shopify into the maker's fortnightly payout and the bank file. Nothing re-keyed. — POP Local - **Reporting & feedback loops** Reviews get requested at the right moment and routed: praise to the website, problems to you first. The Monday report writes itself from the systems you already run, with the three numbers you actually watch on top. — For owner-led teams who want Monday's numbers without building them. — One question after every tour: five stars goes straight to TripAdvisor, anything less to a private form first. The latest reviews sync back to the site every day. — Cheeky Kiwi Travel - **Internal tools & dashboards** None of this is a black box. A dashboard shows what the system handled, what it flagged, and where a human stepped in, and it gives your team the controls to steer it without calling us. — For when the team needs to see and steer all of the above. — Operations dashboards on top of the systems doing the work. — Edtripper and Nilkick — Edtripper's cockpit shows every phone's heartbeat, kiosk state and incidents, and drafts questions with Claude for an educator to approve. Nilkick meters every model call and reports the spend per run. ### Not on the list? The five above cover most of what we're asked to build, not most of what can be built. If it's repeatable, rules-based and eating hours, it's a candidate. The assessment exists to find your version: two weeks mapping the manual work in your business, ranked by what it actually costs you. ### It also looks like this. A missed call gets an SMS with the answer and a booking link before the caller tries your competitor., Supplier invoices matched to purchase orders and pushed into Xero; only the exceptions reach a person., Quote follow-ups that go out on day 3 and day 10, and stop the second the client replies., New-client intake checked, chased and filed before the first appointment., Compliance documents (insurances, certifications, IDs) tracked, chased before they expire, filed., Website enquiries scored and entered into the CRM with a drafted reply waiting., Your own ChatGPT or Claude wired to your systems, so "send Sarah the report on 12 Smith Street" is one sentence and the report goes out., The day's footage read, cut into short edits and posted to your social accounts on a schedule., Stock levels kept in sync between the store, the warehouse sheet and the accounting file., The Monday work-in-progress report written from your job-management system before you're at your desk. ### Assessment. Build. Managed. You don't have to arrive knowing what to buy. That's what step one is for. - **Assessment: decide what's actually worth building.** Two weeks, fixed price. We sit with the people doing the work, map every manual step, and cost each one in hours and risk. You get a ranked list with build prices attached. If the honest answer on something is "don't automate this", the document says that too. The fee is credited against a build started within 90 days. - **Build: one process, end-to-end.** Three to eight weeks. Integrated with what you already run, staged alongside the real thing, your team testing on real data before anything goes live. 30 days of hypercare after. - **Managed: keep it true.** These systems drift: models change, APIs change, your business changes. We watch it and fix it. Improvements are scoped and quoted as they come up. Every month you get a report of what it handled, what it saved, and what it got wrong. Required for the first 12 months. ### Built so you're never stuck with us. - **You own the code and the accounts.** Delivered into your repos and your cloud. No per-task fees. Leave any time. - **Data stays where you say.** Australian-region cloud by default; fully self-hosted when it needs to be. We'll tell you plainly which AI providers see what. - **Human in the loop.** Approvals on anything customer-facing or money-related until you say otherwise. Every action logged; every decision explainable. - **Built for regulated environments.** We've run production systems for an ASX-listed financial services company since 2021. Access control, change review, audit trails aren't add-ons. - **Integrates, doesn't replace.** Your CRM, accounting, booking tool and inbox stay. We connect them. **Works with what you run.** If it has an API, we'll use it. If it doesn't, we've dealt with worse. Plus the in-house system nobody wants to touch. Xero, MYOB, HubSpot, Pipedrive, Google Workspace, Microsoft 365, Calendly, Shopify, Airtable, Slack ### What it costs. Indicative. Fixed price after the assessment. What moves the price within those ranges: - **Automation Opportunity Assessment** $1,500 fixed (2 weeks · credited to your build) - **Automation build** from $8,000 (typically $10–30k per process) - **Managed** from $250/month (monitoring, fixes, monthly report · changes quoted) - **Internal tool / dashboard** typically $8–20k (on top of the systems doing the work) - how many systems we're integrating - whether steps need human approval - how messy the source data is - volume ### Five of them, running. - **Job 01 · Enquiries & bookings** A booking platform on Rezdy with an AI admin that drafts the daily supplier emails from live bookings. Conversion rate up 37% since the rebuild. (https://croissantbaguette.com.au/projects/cheeky-kiwi-travel/) - **Job 05 · Internal tools & dashboards** A venue CMS and fleet cockpit on top of the app, first commit to live phones. Questions drafted by Claude, signed off by an educator, never sent straight to a device. (https://croissantbaguette.com.au/projects/edtripper/) - **Job 04 · Reporting & growth** A full launch pipeline (deep crawl, venue discovery, every draft) costs about a dollar of model spend, metered on every run. The Scout reports back daily, only when there is something new. (https://croissantbaguette.com.au/projects/nilkick/) - **Job 02 · Documents & reports** A custom MCP Joris installed in his own Claude and ChatGPT. One sentence sends a client a property report: generated on the data service, matched to the HubSpot record, emailed. A second worker qualifies every social and website lead into HubSpot. (https://croissantbaguette.com.au/projects/joris/) - **Off menu · Your site, edited by asking** A headless Shopify store the client runs from his own Claude: a skill we wrote plus the Sanity MCP. New landing pages, copy changes, article imports and A/B tests, drafted first and published on his say-so. (https://croissantbaguette.com.au/projects/quitine/) ### Ask away. Twelve things people ask before they book. The full numbers are in our AI automation cost guide. Still one missing? Ask us. - **We already use ChatGPT. What's different?** ChatGPT helps a person do a task. We build a system that does the task itself, connected to your calendar, CRM, inbox and files, and it only asks a person when it should. When it makes sense, we plug our tools into the ChatGPT or Claude you already pay for (an MCP server we host), so you ask in plain English and the system does the work. - **Is this workflow automation?** Partly. Workflow tools like Zapier, Make and n8n move data between apps when something triggers, and we use them where they fit. What we build sits above that: a system that reads the enquiry, document or booking, decides what to do under rules you set, acts across your tools, and asks a person when it should. The workflow is one piece. The system is the thing that runs. - **What if the AI gets it wrong?** It will, sometimes. That's why anything that matters waits for approval, every action is logged, and the managed plan exists. Error rates are measured and reported, not assumed. - **Where does our data go?** Australian-region cloud by default, self-hosted if you need it. We document exactly which providers process what, and we don't train models on your data. - **Do we have to replace our systems?** No. We integrate with what you run. - **What does this mean for our staff?** The system takes the copying, checking and chasing: the parts nobody was hired for. Judgment calls, exceptions and customers stay with your team. In businesses our size the pattern is stopping the next admin hire, not losing people. - **How long until something's live?** Assessment two weeks; first process typically 3–8 weeks after. - **What does it cost to run?** Model/API usage is usually tens to a few hundred dollars a month for an SMB process; the managed fee covers our time. You'll see both on the monthly report. - **Can we get funding for this?** Sometimes. Federal and state programs exist and change often. We keep a current list on the site and will tell you what applies. - **What if we outgrow it / want to leave?** You own everything. We hand over documented, and it keeps running. - **Do you do the "AI strategy" thing?** We do the assessment, which is that with prices attached. Then we build it. - **Who does the work?** Johnny, Alex and Joris. Nobody else. ### Book an assessment. Two weeks. We map what's manual, what it costs you, and what to automate first, with prices attached. Fixed price, credited to your build. ## Free AI visibility and website report https://croissantbaguette.com.au/tools/ai-visibility/ Free report for AU and NZ businesses: are you the answer when customers ask ChatGPT, Claude, Gemini or Google AI? Citation share against two competitors. ### Are you the answer your customers get? When someone asks an AI assistant who does what you do, near you, one paragraph and two or three names come back. We check whether one of them is you, against two competitors, and how your site holds up. Free, for Australian and New Zealand businesses. ### Tell us who you are. We do the asking. Your business, your suburb, what you do and two competitors. We ask ChatGPT, Claude, Gemini and Google AI Overviews the questions your customers ask, check your site's speed and structure, and email you one report with the actual answers. Australian and New Zealand businesses only. Report within the hour. ### One report, four answers. - **Your citation share, against two competitors.** Five questions a real customer would ask, put to four AI engines. How often each of you is named, overall and per engine. - **The actual answers.** Every answer, in full, with the sources it cited. No summary standing in for the evidence. - **Who gets named instead of you.** Every business the engines recommended that isn't you or your two competitors, ranked. - **How your site holds up.** Speed and layout stability from Lighthouse, plus the structure and entity checks that decide whether an AI answer can quote you at all. ### Before you fill it in. Three things people ask. Anything else? Ask us. - **Is it really free?** Yes. Every report costs us real money to run, which is why the form asks for real details and why it is for Australian and New Zealand businesses only. No card, no call required. - **How long does it take?** Usually under 15 minutes, occasionally up to an hour if one of the engines is slow or a person needs to check the request first. - **What happens with my details?** We email you the report and may follow up once to ask if it was useful. Your name and email never go to any AI service: only your business name, website, suburb and services do, and those are public already. ### Book an assessment. Two weeks. We map what's manual, what it costs you, and what to automate first, with prices attached. Fixed price, credited to your build. ## SEO & AI Visibility https://croissantbaguette.com.au/ai-visibility-seo/ Are you cited when customers ask ChatGPT, Claude or Google AI? Free report, $950 audit, monthly citation reporting. Sydney studio that builds the fix. **Service:** AI Search Visibility (GEO) and Technical SEO (Generative engine optimisation) We measure whether your business is the answer customers get from ChatGPT, Claude, Gemini and Google AI Overviews, fix what stops you, and report citation share every month. - **Free AI visibility and website report** — Free report for Australian and New Zealand businesses. Citation check across ChatGPT, Claude, Gemini and Google AI Overviews, plus website speed and structure. Emailed within the hour. - **AI Visibility Audit** from A$950 — 30–50 real customer questions, all engines, competitor share, technical SEO/GEO fixes and a 90-day plan. Credited against the programme. - **GEO & SEO programme** — Technical fixes, content infrastructure, corroboration and monthly citation-share reporting. From $2,000 a month, 3-month minimum. ### Be the answer, not the tenth blue link. Your customers now ask AI before they call anyone. We measure whether you're the answer they get, fix what stops you, and report the number every month. ### Search moved. Most Australian sites didn't. AI Overviews now appear on roughly half of Google searches. ChatGPT passed 900 million weekly users this year. When someone asks "who does [your service] in [your suburb]" they increasingly get one paragraph and two or three names, not a page of links. If you're not in that paragraph, you're invisible in a way rankings never made you. Getting cited is a different game from ranking. AI answers reward pages that state facts plainly, answer a specific question, carry clean structured data, and are corroborated elsewhere. Marketing prose loses. Clear pages win. ### Are you the answer? Find out, free. Tell us your business, your suburb, what you do and two competitors. We ask ChatGPT, Claude, Gemini and Google AI Overviews the questions your customers ask, check your site's speed and structure, and email you one report with the actual answers. Australian and New Zealand businesses only. Report within the hour. ### Audit, then programme. The free report tells you if there's a problem. The audit tells you why. The programme fixes it. - **AI Visibility Audit.** $950 The full version of the free report: 30–50 real customer questions, all engines, competitor share, why you're missing (entity, structure, corroboration, technical), technical SEO/GEO fixes for your site, and a 90-day plan. Credited against the programme. - **GEO & SEO programme.** from $2,000 We fix the technical layer, build the content infrastructure (the specific pages that answer the specific questions), get you corroborated where AI models look, and report citation share and organic performance monthly. If the number doesn't move, we say so. ### What we measure. - **Citation share** How often you're named across the question set, per engine, vs competitors. Tracked monthly. - **Traffic and enquiries** Organic and AI-referred traffic and enquiries. - **Technical** Structured data validity, indexation, Core Web Vitals. - **ChatGPT** 44 - **Claude** 38 - **Gemini** 27 - **Google AI Overviews** 31 Competitor A · 41%, Competitor B · 22%, Structured data · valid, CWV · pass ### Why us for this. - **We build the pages, not just the recommendations.** Structured data, semantic HTML, sub-second load: that's how every site we ship is built already. Most GEO agencies hand you a doc; we hand you the fix, deployed. - **We've built a content-and-search system before.** Nilkick, our own product, ships 57 guides from one content source, each with a markdown twin for AI crawlers, llms.txt and structured data. Its scanner reads the same signals on other people's sites: 27 launch-readiness rules and 19 agent-readiness checks, no model deciding the number. - **We prove it on ourselves.** This site is the test bed; our own citation share is on the report we send you. - **Marketing is in the name.** C&B Marketing Pty Ltd: the team came from demand generation before it came to code. ### What it costs. - **Free AI visibility and website report** free (emailed within the hour · shareable link) - **AI Visibility Audit** $950 fixed (10 days · credited to programme) - **GEO & SEO programme** from $2,000/mo (3-months min) - **Technical SEO fixes only** from $2,000 (existing sites · typically $2–6k) ### Ask away. Five things people ask before they book. Still one missing? Ask us. - **Is this SEO?** It includes technical SEO. But the goal is being cited in the answer, which rewards different things than ranking: clarity, structure, corroboration. - **Can you guarantee we'll be cited?** No, and anyone who does is guessing. We can guarantee you'll know the number every month and what we did to move it. - **We already have an SEO agency.** Fine. We can do the audit and the technical layer and hand them the content plan. - **Do we need a new website?** Usually not. Sometimes the site is the reason you're invisible; the audit will say. - **How long before it moves?** Technical fixes: weeks. Citation share: usually 2–4 months of consistent, specific content. ### Get the free report. Your business, your suburb, two competitors. We show you who gets named in AI answers, with the actual answers, and how your site holds up. ## Contact https://croissantbaguette.com.au/contact/ Tell us what's slow, manual or invisible, or what you need built. Straight answer within a business day. Sydney studio: websites, apps, AI systems, growth. ### Tell us what's slow, manual, or invisible. Or what you need built. Two lines is enough. We'll come back within a business day with a straight answer and a next step. No forms that pre-qualify you to death. Reply within one business day. ## eCommerce https://croissantbaguette.com.au/web-development/ecommerce/ Custom Shopify themes from $9k, headless storefronts from $15k. Sub-second, brand-true, editable by your team, yours to keep. Built in Sydney. **Service:** Ecommerce development (Web development) Headless storefronts and custom Shopify themes on Shopify checkout, designed and built in Sydney and owned by the client (storefront code, content and store). - **Headless storefront** from A$15000 — Custom storefront on a modern JS framework, Shopify checkout, Sanity content. From $15,000, typically $18–35k. - **Custom Shopify theme** from A$9000 — Built from a clean base, launched in weeks, editable in the theme editor. From $9,000, typically $9–16k. - **Care plan** from A$250 — Monitoring, updates, small changes and a monthly performance snapshot. From $250 a month. ### Shopify under the hood. A store that's actually yours on top. Two ways in: a headless storefront when speed, brand and freedom matter, or a custom Shopify theme when time to launch does. Both fast, both editable by your team, both built in Sydney. **Spec ticket** Fixed quote after one call. - **Runs on** Shopify checkout, always - **Storefront** headless · or custom theme - **Typical build** 3–10 weeks - **After launch** care plan optional - **<1s every page** — POP Local · 10k SKUs, from 20 seconds - **33% conversion rate** — Made Quiet · headless Shopify, built in six days - **AU+NZ one theme, every store** — Constellation Brands - **Double digit growth in sales** — Quitine · since moving off the theme ### Most stores are slowest where it costs the most. Product and collection pages carry the weight: apps stacked on a theme, a script from every tool you ever trialled, images nobody resized. POP Local's category pages were the slowest thing on the site, and those are the pages people shop from. Karis Life's and Kea Health's WooCommerce stores fell over under the traffic they'd worked to earn. A store that's slow at the moment of buying is paying for visitors it can't convert. ### Two ways to build it. We'll recommend one after a 20-minute call. Both keep Shopify's checkout, and both are yours. - **Headless storefront** speed and brand are the point, you want custom features a theme can't do, you have a lot of content around the products, or the store is one part of a bigger site. — POP Local (343 makers, under a second on every page, moved from Square with the point of sale) · Made Quiet (built in six working days after a week of design) · Quitine (off a Shopify theme, now edited by asking Claude) - A custom storefront on a modern JS framework, pre-rendered and served from the edge. Sub-second product pages. - Shopify keeps cart, checkout, payments, inventory and orders. Nothing customers trust gets rebuilt. - Content in Sanity with live preview; products stay in Shopify. No overlap, no confusion. - No theme limits on layout, search or filtering. If you can draw it, we can build it. - **Custom Shopify theme** you want to launch in weeks, your team lives in Shopify admin, and the design fits inside what themes do well. — Constellation Brands: one enterprise theme across their Australian and New Zealand stores, dozens of custom sections, all editable by their team. - Built from a clean base, not a marketplace theme with your logo on it. - Sections and blocks your team rearranges in the theme editor without calling us. - Only the apps that earn their place. Every app is another script on your product page. - Multi-region and multi-currency through Shopify Markets when you need it. ### Headless or theme? Go headless when you want freedom: any layout, search and filtering built your way, content around the products, sub-second pages. Pick a theme when launching in weeks inside Shopify admin matters more. - **Speed** Sub-second by design | Fast for a theme; each app added costs a little - **Time to launch** 6–10 weeks | 3–6 weeks - **Design freedom** Anything | Within what sections and Liquid allow, which is a lot - **Editing** Sanity for content, Shopify for products | Shopify's theme editor, all in one admin - **Apps and integrations** Shopify apps for checkout and back office; storefront features built by us | The whole Shopify app store, drop-in - **Best for** Brand-led stores, custom features, content-heavy catalogues, stores inside a bigger site | Teams who want everything in one admin and a launch in weeks ### Whichever way, you get this. - **Shopify checkout.** Payments, fraud, tax, shipping and Shop Pay, handled by the platform that processes billions. We don't rebuild the part customers trust. - **Speed measured, not promised.** Core Web Vitals checked before launch and in every care-plan report. - **Built to be found.** Product structured data, clean collections, fast pages: what Google Shopping and AI shopping answers reward. - **Migration handled.** Products, customers, orders and URLs moved across; old links redirected so nothing you've earned is lost. - **Connected to the back office.** Accounting, shipping, email, reviews and point of sale. - **Yours to keep.** Theme code in your repository, store in your Shopify account. Any Shopify developer can take over. ### Stores we've shipped. - An enterprise Shopify theme across Australia and New Zealand, cleared to US corporate standards. (https://croissantbaguette.com.au/projects/constellation-brands/) - 343 Australian makers, under a second on every page, and a portal that pays them. (https://croissantbaguette.com.au/projects/pop/) - A headless store built in six working days after a week of design. First sale in week one, 33% conversion rate. (https://croissantbaguette.com.au/projects/made-quiet/) - A Shopify theme replaced by a headless store the client runs by asking Claude. Double-digit sales growth since. (https://croissantbaguette.com.au/projects/quitine/) - WordPress for content and Shopify for sales became one store. (https://croissantbaguette.com.au/projects/growing-spaces/) - A WooCommerce store that kept crashing, replaced with one that doesn't. (https://croissantbaguette.com.au/projects/karis-life/) - A store that buckled under demand, rebuilt to keep up with it. (https://croissantbaguette.com.au/projects/kea-health/) ### Scope. Design. Build. Launch. - **Scope.** A 20-minute call, then a fixed quote. Headless or theme, and why, in writing. - **Design.** Home, collection, product and cart designed on your real products, not placeholders. - **Build and migrate.** Storefront built, apps connected, catalogue and customers moved across, checkout tested end to end. - **Launch and after.** Redirects, analytics, Search Console, and a month of watching the numbers. Then a care plan if you want one. ### Ranges. Fixed quote after a scoping call. No calculator; the interesting variables are never on a form. What moves the price within those ranges: - **Headless storefront** from $15,000 - **Custom Shopify theme** from $9,000 - **Care plan** from $250/month (monitoring, updates, small changes, monthly performance snapshot) - catalogue size and how much moves across - number of custom sections or page templates - integrations (accounting, shipping, email, reviews, POS) - multi-region or multi-currency ### Ask away. Six things people ask before they book. The full numbers are in our Shopify cost guide. Still one missing? Ask us. - **Do we have to use Shopify?** No. We've built on BigCommerce, and custom setups on Stripe and Square. Shopify is the default because it's the one we'd pick for ourselves: checkout, payments and fraud handled by a platform that does it at scale. - **We're on WooCommerce, Squarespace or Wix. Can you move us?** Yes. Products, customers, orders and URLs come across, with redirects so you keep your rankings. Karis Life and Kea Health both came from WooCommerce stores that couldn't take the load. - **Headless or theme?** Headless when you want freedom over layout, search, filtering and custom features, when there is a lot of content around the products, or when the store lives inside a bigger site. A theme when launching in weeks inside Shopify admin matters more. We'll tell you which on the first call. - **Can we start with a theme and go headless later?** Yes. The store, products, orders and customers stay where they are. Headless replaces the storefront, not the shop. - **Will our team be able to update it?** Yes. Theme builds use Shopify's editor. Headless builds use Sanity for content with live preview, and Shopify admin for products, the way your team already works. - **Why is our current store slow?** Usually apps. Each one adds scripts to every page. We audit what's installed, keep what earns its place, and rebuild the rest into the storefront. Then we measure it. - **Websites.** Pre-rendered and sub-second, or WordPress built properly. - **Web & mobile apps.** Portals, bookings, dashboards, and the app your team carries. - **AI Automation.** Orders, enquiries and stock levels handled without a person re-keying them. ### Tell us what you sell and where it's slow. Two lines is enough. You'll get a headless-or-theme recommendation and a fixed quote within a business day of the call. ## Guides https://croissantbaguette.com.au/guides/ What a website, a Shopify store and AI automation actually cost in Australia in 2026. Real bands, real running costs, and real projects itemised line by line. ### Straight answers about money. What things cost in Australia, written for the person paying rather than for search engines. Every guide has the market's bands, what moves the price, what we charge, and at least one real project itemised line by line. **Quick answer** Every figure on these pages excludes GST. - **Website** $5,000–15,000 - **Store** $56 a month, plus the build - **Automation** $5,000–15,000 a process - **To run** $0–200 a month ### Because "it depends" is not an answer. Ask what a website costs and you get a range so wide it tells you nothing, or a form that promises a quote and delivers a phone call. We publish our prices on every service page, so writing these was mostly a matter of showing our working: what each band actually buys, what moves a number within a range, what the thing costs to run once it exists, and what real projects came to, line by line. Where a figure comes from somebody else we say who and when. Where it comes from our own books we round it and say so. Where we cannot source it, it is not here. ### Three of them, so far. - **Website costs.** What a website costs in Australia in 2026, by band, plus the monthly bill nobody quotes properly and two real builds itemised. - **Shopify costs.** What Shopify charges in Australian dollars, what a store costs to build, the app bill nobody budgets for, and one real store itemised. - **AI automation costs.** What it costs to take one process off your team, why the hours matter more than the tiers, and three systems itemised including our own. ### Or skip the reading. Tell us what you need in two lines and we will give you a number and a date, or tell you honestly that we are not the right people for it. ## How much does a Shopify store cost in Australia? https://croissantbaguette.com.au/guides/how-much-does-a-shopify-store-cost-in-australia/ Shopify costs $56 a month in Australia, plus $1,500 to $35,000 to build. Plan prices, card rates, the app bill, build bands, and one real store itemised. ### How much does a Shopify store cost in Australia? There are three bills, and most guides only count two. Shopify charges $56 a month on Basic. A build runs from $1,500 for a fitted theme to $35,000 for a headless storefront. Then there is the app stack, which is the one that surprises people. **Quick answer** Shopify prices are AUD list. Ours exclude GST. - **Shopify** $56 a month - **Theme build** $1,500–16,000 - **Headless** $15,000–35,000 - **Apps** $0–500 a month - **Card rate** 1.75% + 30c ### $56 a month to Shopify, and $1,500 to $35,000 to build it. Shopify's Basic plan in Australia is $56 a month, or $42 if you pay yearly, and it takes 1.75% plus 30 cents from a domestic card sale. That part is fixed, published, and the same for everyone. What varies is the storefront on top of it: $1,500 to $3,500 for a freelancer to set up a bought theme, $9,000 to $16,000 for a custom theme built from a clean base, and $15,000 to $35,000 for a headless storefront where Shopify keeps the checkout and everything a customer sees is yours. The third bill is apps, and it is the one nobody puts in the budget. Reviews, subscriptions, bundles, upsells, a page builder: $10 to $200 a month each, most of them adding a script to every product page. Work out what the store will run on before you decide what it costs to build. ### What Shopify itself charges. Australian list prices, September 2026. This is the part of the bill that has nothing to do with who builds your store. Australian list prices from shopify.com/au/pricing, read on 10 September 2026. Card rates are for a domestic card through Shopify Payments; international and American Express cards cost more, and using a third-party gateway instead adds 0.2% to 2% on top depending on the plan. - **Basic · $42 billed yearly** $56 One storefront, the full online store, up to 10 inventory locations. Where nearly every new Australian store starts, and where most stay longer than they expect to. — 1.75% + 30c, so $2.05. - **Grow · $114 billed yearly** $149 Staff accounts, better reporting, lower card rates. The upgrade you make when the card-rate saving covers the difference, not before. — 1.6% + 30c, so $1.90. - **Advanced · $431 billed yearly** $575 Custom reporting, deeper international shipping, more staff accounts. Rarely earns its place under seven figures of revenue. — 1.4% + 30c, so $1.70. - **Plus** from $3,700 Enterprise. Checkout customisation, more storefronts, higher API limits, an account team. — Negotiated. Third-party gateway fee drops to 0.2%. ### The app bill is the one nobody budgets for. Reviews, subscriptions, bundles, upsells, a page builder, a currency switcher, a loyalty scheme. Apps run from about $10 to $200 a month each, so a store on the $56 plan running eight of them is usually paying more for its apps than for Shopify. It compounds twice, in money and in speed, because most of them add a script to every product page. POP Local's category pages, the pages people actually shop from, were the slowest thing on the site before we rebuilt them. Every app is a decision you are making about your product page load time, whether or not anyone framed it that way. ### What the storefront costs. What an Australian buyer pays in 2026 to get a store built, on top of Shopify's own subscription. Our own numbers are further down. Figures exclude GST and exclude Shopify's own subscription and card fees, which you pay to Shopify. - **Do it yourself** $0–500 A free or paid theme you set up yourself. Shopify's own themes are genuinely good, and a tidy store on Dawn beats a bad custom one every time. — A first product, a small range, or finding out whether people will buy at all. - **A theme, set up for you** $1,500–3,500 A freelancer installs a paid theme, loads your products, connects payments and shipping and hands it over. Two to three weeks. — A launch with a date on it and a catalogue small enough to load by hand. - **A theme, customised** $3,500–9,000 A bought theme reworked to your brand. Custom sections, a homepage that is yours rather than the theme's, collection and product pages that match the rest of your world. — Brands that need to look distinct without leaving Shopify admin. - **A custom theme** $9,000–16,000 Built from a clean base instead of a marketplace theme. Sections and blocks your team rearranges without calling anyone, and only the apps that earn their place. — Teams who want everything in one admin and a launch in weeks. - **A headless storefront** $15,000–35,000 A custom storefront on a modern framework, with Shopify keeping cart, checkout, payments and inventory. Sub-second pages, and no theme limits on layout, search or filtering. — Brand-led stores, large catalogues, content-heavy sites, or a store that lives inside a bigger site. ### What actually moves the number. - **How many products, and how messy they are.** Forty products with clean data is an afternoon. Ten thousand with variants, options, and images coming from three different places is a project of its own. POP Local has 10,000 across 343 makers. - **Whether the design fits inside what a theme does.** This is the line between a $9,000 build and a $20,000 one. Themes do a great deal well; they do not do arbitrary layout, custom search, or filtering that behaves the way you drew it. - **What the store has to talk to.** Accounting, shipping, email, reviews, a point of sale, a warehouse spreadsheet. Each integration is real work, and each one can wait a week on the other side's API. - **Whether you are moving, and from where.** Products, customers, orders and every old URL redirected. WooCommerce and Squarespace exports are rarely as clean as they look. - **Multi-region and multi-currency.** One store in Australia is one build. Australia and New Zealand with different catalogues, pricing and shipping is meaningfully more, even through Shopify Markets. - **How many people will run it afterwards.** A store one person updates is not the store six people update, with roles, drafts and someone approving before it goes live. ### One real store, line by line. Made Quiet makes premium Australian sports caps. There was a name, a range, a clear idea of the brand, and no store, with a first drop already booked for February. A week of design, then six working days from first commit to a live store. It sold in its first week, a third of visitors who arrive go on to buy, and seven months in the total is closing on six figures. Here is what it cost, and what it costs to run. Published with Made Quiet's permission. Excludes GST. Built January 2026. This one lands below our own headless floor of $15,000, which is worth explaining rather than hiding. It was January 2026, one small range with no catalogue to migrate, one set of templates, two integrations, and a client whose brand and copy were ready before we started. A store with four hundred products, a WooCommerce migration and an accounting system to stay in sync with is a different build at a different price, and that is what the floor is for. We would quote this one higher today. - **Client** New brand, first store - **Scope** Headless storefront, Sanity content, Klaviyo, reviews - **Built in** 7 days design, 6 days build. January 2026 - **Live at** madequiet.co - **Design** $3,000 7 days. Home, collection, product and cart, on their real products - **Development** $9,000 6 working days. SvelteKit on Cloudflare Workers, Shopify Storefront API, own cart - **Total** $12,000 ex GST **Costs to run** No care plan. Changes are quoted and billed by the hour when they want them. - **Shopify** A$56 Basic plan. Products, cart, checkout and payments - **Storefront hosting** $0 Cloudflare Workers, their own account - **Sanity, Klaviyo, Judge.me** $0 content, email capture, reviews and star ratings ### What we charge. Fixed quote after a scoping call. Shopify's subscription and card fees are yours and go to Shopify; they are not part of what we charge. - **Headless storefront** from $15,000 (Shopify checkout, storefront built by us) - **Custom Shopify theme** from $9,000 (built from a clean base, not a marketplace theme) - **Care plan** from $250/month (monitoring, updates, small changes, monthly performance snapshot) ### Ask away. Eight things people ask before they spend the money. Still one missing? Ask us. - **How much does it cost to pay someone to build a Shopify store?** In Australia in 2026, $1,500 to $3,500 for a freelancer to set up a bought theme, $3,500 to $9,000 for that theme customised properly, $9,000 to $16,000 for a custom theme built from a clean base, and $15,000 to $35,000 for a headless storefront. All of it sits on top of Shopify's own subscription, which you pay separately. - **How much does Shopify take from a $100 sale?** On the Basic plan in Australia, $2.05, which is 1.75% plus 30 cents through Shopify Payments on a domestic card. Grow takes $1.90 and Advanced takes $1.70. International and American Express cards cost more, and using a third-party gateway instead of Shopify Payments adds another 0.2% to 2% depending on your plan. - **How much does a Shopify store cost per month?** $56 a month for Basic in Australia, or $42 if you pay yearly. Apps are where the number really moves, at roughly $10 to $200 a month each, and email marketing adds $20 to $100. A headless storefront hosted on Cloudflare adds close to nothing. Budget $100 to $400 a month all in for a small store that is running properly. - **Is Shopify worth it in Australia?** For almost every store below enterprise, yes. What you are buying is checkout, payments, fraud, tax and shipping being somebody else's problem, at a scale you could not build for yourself. The parts worth arguing about are the storefront and the app stack, and those are exactly the parts you can own. - **Do we have to use Shopify?** No. We have built on BigCommerce, and custom setups on Stripe and Square. Shopify is the default because it is the one we would pick for ourselves: the part customers trust is handled by a platform that does it at scale, and we do not rebuild that. - **Does a headless store cost more to run?** It costs more to build and less to run. The storefront is pre-rendered and served from a CDN, so hosting is close to nothing and pages load in under a second. A theme launches in weeks and keeps everything in one admin. We will tell you which one you want on the first call, and it is not always ours. - **What does it cost to move from WooCommerce, Squarespace or Wix?** The migration is usually $1,000 to $4,000 on top of the build. Products, customers, orders and every old URL redirected, so you keep the rankings you have earned. Karis Life and Kea Health both came across from WooCommerce stores that could not take the load. - **Why is our current store slow?** Usually apps. Each one adds scripts to every page, and product and collection pages carry the most of them. We audit what is installed, keep what earns its place, rebuild the rest into the storefront, and then measure it rather than claiming it. - **Website costs.** What a website costs in Australia, by band, with a real receipt. - **AI automation costs.** What it costs to take one process off your team. - **eCommerce.** Headless or a custom theme. What we build, and how. ### Tell us what you sell. Two lines is enough. You get a headless-or-theme recommendation and a fixed quote within a business day of the call, and a straight answer if a fitted theme is really what you need. ## How much does a website cost in Australia? https://croissantbaguette.com.au/guides/how-much-does-a-website-cost-in-australia/ Most Australian businesses pay $5,000 to $15,000 for a website designed and built for them in 2026. The bands, the drivers, running costs, and two real builds itemised line by line. ### How much does a website cost in Australia? Most Australian businesses pay $5,000 to $15,000 for a site designed and built for them, and close to nothing to run it if it was built right. Here are the bands, what moves the price, what we charge, and two real builds itemised line by line. **Quick answer** Every figure on this page excludes GST. - **DIY** $0–1,500 a year - **Template** $2,000–5,000 - **Designed** $5,000–15,000 - **Custom** $15,000–40,000 - **To run** $0–150 a month ### $5,000 to $15,000, if someone designs and builds it for you. In Australia in 2026, a website designed and built for a business by a studio or an agency costs $5,000 to $15,000. Below that, from about $2,000, you are paying someone to fit a template. Above it, from about $15,000, you are paying for the site to do a job: bookings, a portal, a large catalogue, systems it has to talk to. Do it yourself with a website builder and the cash cost is $0 to $1,500 a year, and the real cost is your own time. Running it is a separate bill, and a much smaller one than most guides suggest. A site built ahead of time and served from a CDN costs close to nothing a month, because nothing runs when a visitor arrives. A site that executes software on every request, WordPress being the common case, costs $30 to $100 a month to host and needs someone to keep it patched. Neither number includes marketing, which is where the $500 to $5,000 a month retainers quoted elsewhere actually live. ### What each band buys. Five ways to get a website, in the order people usually consider them. These are what an Australian buyer pays in 2026, not what we charge; our own numbers are further down. Figures exclude GST. We publish our own numbers below: we start at $9,000 and most builds land between $9,000 and $18,000, so we sit at the top of the third band and into the fourth. - **Do it yourself** $0–1,500 a year A Wix, Squarespace or Shopify template you set up yourself. Subscription, domain, and a paid theme if you want one. Every hour of the work is yours. — Testing an idea, a side business, or anyone whose customers will never judge them on a website. - **A template, fitted** $2,000–5,000 A freelancer or a small shop puts your logo, colours and words into a bought theme. Five to ten pages, live in two or three weeks. It looks like what it is. — A first site when cash is the binding constraint, or a business that needs to exist online and nothing more. - **Designed and built** $5,000–15,000 Designed from nothing for your business, built on a CMS your team can actually use, structured for search and built to load fast. Eight to twenty pages, four to eight weeks. — Established businesses where the site is the first impression, and one extra enquiry a month pays for the whole thing. - **Built to do a job** $15,000–40,000 All of the above, plus the site doing work: bookings, a client portal, payments, a CRM it stays in sync with, a catalogue with real filters, or several hundred pages migrated off an old platform. — Businesses whose website does the work rather than describing it. - **Software with a website on the front** $40,000+ A custom application. Marketplaces, multi-site platforms, anything with logins, roles, permissions and a data model of its own. — Product businesses, and organisations running several brands off one codebase. ### The cheap one is rarely the cheap one. The number that decides this is not what the site costs to build. It is what it costs over four years. A $3,000 site that takes seven seconds to load, that nobody on your team can edit, and that needs a developer booked in for every change, spends the rest of its life costing you money in ways that never appear on an invoice. We have measured both ends of it. Wingboy's site took 27 seconds to load before we rebuilt it and 0.7 seconds after, with the same content. NIODA's WordPress site took ten seconds, and exactly one person in the organisation could edit it; the rebuild made that four, and they have published 391 changes themselves since. Neither business was badly served by a platform. They were badly served by a site nobody owned. ### What actually moves the number. - **Templates, not pages.** Twenty pages built on five templates costs less than eight pages that are all different. When a quote asks how many pages you need, this is the thing it is really asking. — The two receipts below are a marketing site and a 958-venue directory, and they came in four hundred dollars apart. - **Where the words and pictures come from.** Writing the copy, shooting the photography, or moving three hundred pages off an old site are each their own line on the quote. Content is the most common reason a build runs late and over. - **What it has to talk to.** A contact form is free. A booking system, a CRM, a payment gateway, an inventory file, or an in-house system with no API, each add days and each can add a week if the other side is slow. - **Who is allowed to edit it.** One person changing a headline is a different build from twelve people across three offices, with different permissions and an approval step before anything publishes. - **What happens to the old URLs.** A site that already ranks needs every old address mapped to a new one before launch day, or you buy a redesign and pay for it again in lost traffic. - **Whether speed is a requirement or a wish.** Building for sub-second loads is a decision made in the first week, about how pages are rendered and where they are served from. Retrofitting it afterwards is a second project. ### Two real builds, line by line. Every guide to this question, ours included, quotes a range. Here is what two builds actually came to, and the pair is worth more than either on its own. Joris is a buyer's agent who started in Newcastle with no site, no rankings and established agents to compete with. He needed to look established and be found: service pages, success stories and resources written for the searches Newcastle buyers actually make, on a domain with no history. It loads in under half a second. Published with Joris's permission. Excludes GST. - **Client** Buyer's agent, Newcastle. New business, no site - **Scope** Marketing site, CloudCannon CMS, structured data - **Built in** 14 days design, 3 weeks build - **Live at** joris.com.au - **Design** $3,600 14 days, by a partner studio we work with - **Development** $10,000 3 weeks. Pre-rendered SvelteKit on Cloudflare, edited in CloudCannon - **Total** $13,600 ex GST **Costs to run** Everything else sits inside a free tier. His accounts, his name on the bills. - **Hosting** $0 Cloudflare, his own account - **CMS** US$10 CloudCannon on our partner rate. List pricing is higher - **Care plan** $350 covers this site and the AI systems we also built him ### The same money, 958 pages. Edtripper's website had to become a directory teachers would actually use: 958 venues across every state and territory, filtered by subject, year group and attraction type, with the whole thing publishable by a team that does not include a developer. A marketing site and a 958-venue directory, four hundred dollars apart. That is the first driver on the list doing its work. You are paying for templates, not for pages, and a directory is one set of templates with a great deal of content behind it. Published with Edtripper's permission. Excludes GST. Both sit in the third band. Neither number includes the AI systems we later built for the same two clients, which are itemised in the AI automation cost guide. - **Client** Founder-led education business, Sydney - **Scope** Directory site, Sanity CMS, 958 venue listings - **Built in** 7 days design, 14 days build - **Live at** edtripper.com - **Design** $4,000 7 days. One set of templates, used 958 times - **Development** $10,000 14 days. SvelteKit on Cloudflare Workers, state-aware queries, filters - **Total** $14,000 ex GST **Costs to run** Hosting, the CMS and email are all free at this size. The care plan is a choice, not a requirement. - **Hosting** $0 Cloudflare Workers, their own account - **CMS** $0 Sanity, inside the free tier at 958 venues - **Email** $0 Resend, on our business account at no charge to them - **Care plan** $250 the website side. The phone fleet is billed separately ### What it costs to keep alive. Building it happens once. This is the bill that arrives every month, and it turns almost entirely on whether the site runs software when somebody visits it. Excludes GST. Australian list prices as at September 2026. - **Hosting a pre-rendered site** $0 Every page is built ahead of time and served from a CDN, so nothing executes when a visitor arrives and there is nothing to pay for. Both builds itemised above host for nothing, and so does this page you are reading. - **Managed WordPress hosting** $30–100 A server running PHP and a database on every request, with backups and a CDN in front. Cheaper hosting exists; this is what it costs to be fast and to stay up. - **A.com.au domain** $2–4 About $20 to $40 a year. The one bill nobody should ever let lapse. - **The CMS** $0–200 Free at the sizes most businesses run at. Edtripper's Sanity costs nothing at 958 venues; Joris is on CloudCannon at US$10 a month through our partner rate, and list pricing is higher. Paid plans start when you add editors, roles, or a lot of assets. - **A care plan** $250–350 Monitoring, updates, small changes and a monthly performance snapshot. Edtripper pays $250 for the website side, Joris $350 for the site plus the AI systems, and Made Quiet pays nothing and buys changes by the hour. Optional on a pre-rendered site, which runs perfectly well without anyone. Not really optional on WordPress, where somebody has to patch it every month. - **A growth retainer** $500–5,000 SEO, content and ads. Often worth having, and a completely different purchase from keeping a website online. Guides that fold it into "running costs" make a website look five times more expensive than it is. ### What we charge. Fixed quote after a 20-minute call. No calculator; the variables that actually move the price are never on a form. - **Website build** from $9,000 (pre-rendered or WordPress, designed from scratch) - **Content migration** quoted (from an existing site) - **Care plan** from $250/month (monitoring, updates, small changes, monthly performance snapshot) ### Ask away. Eight things people ask before they spend the money. Still one missing? Ask us. - **What is a fair price to pay for a website?** For a small business site designed and built for you, $5,000 to $15,000 in Australia in 2026. Under $2,000 you are buying a fitted template, which is a fair price for what that is. Over $15,000, someone should be able to point at the thing the extra buys: an integration, a portal, a migration, a catalogue. If nobody can name it, it is not a fair price. - **Is it cheaper to build a website myself?** In cash, yes: $0 to $1,500 a year on Wix, Squarespace or Shopify. In time, usually not. A first build takes most owners 40 to 80 hours, and what you get is a template, which is fine right up until the site has to do something templates do not do. Build it yourself while the business is finding its feet, and pay someone once the site starts costing you enquiries. - **How much does a website cost per month?** Close to nothing if it is pre-rendered, because nothing runs when a visitor arrives. Of the two builds itemised on this page, one runs on $0 a month and the other on US$10, and that ten dollars is the CMS rather than the hosting. Budget $30 to $100 a month for managed WordPress hosting instead, plus about $20 to $40 a year for a.com.au domain either way. A care plan starts at $250 a month and is optional on a pre-rendered site. The $500 to $5,000 a month quoted in other guides is a marketing retainer, not the cost of keeping a website online. - **How much should a small business website cost?** Between $5,000 and $15,000 across the Australian market, and between $9,000 and $18,000 with us. The floor is set by the work that does not scale down: structure, design, a CMS, content, testing and launch. A five-page site and a fifteen-page site cost far more alike than people expect. - **How long does it take to build a website?** Four to eight weeks for a designed and built site, and most of that is waiting on content rather than code. We quote a launch date after a 20-minute call and hold it. - **Do I have to pay for hosting and maintenance forever?** You pay for hosting, which on a pre-rendered site is close to nothing. Maintenance is a choice. A pre-rendered site has no plugins, no database and no login page, so there is nothing to patch and it will still be running in three years without us. A WordPress site needs somebody every month, whether that is us, another agency, or you. - **What is the cheapest way to host a website in Australia?** Free, if the site is pre-rendered. Cloudflare, Netlify and Vercel all serve static pages at no cost at the volumes a business site does, from servers in Sydney. Shared hosting at $5 a month is more expensive than free and considerably slower. - **Why do two quotes for the same website differ by $20,000?** Because "website" describes about twenty different products. Ask each quote the same four questions: what happens to our old URLs, who can edit the site afterwards, what will it load in, and what happens if the person who built it disappears. The gap usually closes into something explainable. When it does not, you have learned something. - **Shopify store costs.** What a store costs to build, and what Shopify itself takes. - **AI automation costs.** What it costs to take one process off your team. - **Websites.** What we build and how. Pre-rendered, or WordPress done properly. ### Tell us what the site needs to do. Two lines is enough. You get a platform recommendation and a fixed quote within a business day of the call, and a straight answer if your budget and your brief do not meet. ## How much does AI automation cost in Australia? https://croissantbaguette.com.au/guides/how-much-does-ai-automation-cost-in-australia/ Automating one business process end to end costs $5,000 to $15,000 in Australia, plus $20 to $200 a month to run. The bands, the drivers, and three real systems itemised, from $5,000 to $36,000. ### How much does AI automation cost in Australia? Taking one business process off your team end to end costs $5,000 to $15,000 to build and $20 to $200 a month to run. The number that decides whether it is worth it is not the build price. It is how many hours it takes back, and what those hours cost you. **Quick answer** Every figure on this page excludes GST. - **Off the shelf** $0–100 a month - **One workflow** $1,500–5,000 - **One process** $5,000–15,000 - **Harder ones** $15,000–50,000 - **To run** $20–200 a month ### $5,000 to $15,000 for one process, end to end. In Australia in 2026, having one defined business process automated end to end costs $5,000 to $15,000, and $20 to $200 a month in model and API usage to keep running. Below that, from about $1,500, you are buying a workflow: two or three tools wired together, which works until something unusual arrives. Above $15,000 you are usually paying for the hard parts, which are approvals, audit trails, systems with no API, and the operations dashboard your team steers it from. Almost every guide to this question is written for people who want to start an AI automation agency rather than buy from one, which is why the numbers you find are all over the place. Here is the buyer's version, with our own prices in it, and two systems we actually built. ### Price it against the hours, not the software. Every pricing guide for this work sorts by complexity: a basic tier, a standard tier, an enterprise tier. It is the wrong axis. The only number that decides whether an automation is worth building is how many hours it takes off your team, and what those hours actually cost you. The ABS put average weekly ordinary time earnings for full-time adults at $2,083.70 in May 2026, which is about $55 an hour before superannuation, payroll tax, leave and everything else that sits on top of a wage. Call a loaded admin hour $65 to $75. A system that removes fifteen hours a week is taking roughly 780 hours a year back, which is $50,000 to $58,000 of loaded cost, and materially more when the person doing it is billable. Against that, an $8,000 build with a $250 a month managed fee is a question about payback period, not about price. It cuts the other way too. If the honest answer is that a process eats two hours a month, do not automate it. Finding that out before anyone spends anything is exactly what the assessment is for. ### What each band buys. What an Australian buyer pays in 2026 to have work taken off their team, from the cheapest thing that helps to the point where it becomes software development. Our own numbers are further down. Figures exclude GST and exclude model and API usage, which is a monthly bill and is set out below. We start at $8,000 and most builds land between $10,000 and $30,000 per process. The three receipts further down came to $5,000, $15,000 and $36,000, and each one explains why it sits where it does. - **Off-the-shelf tools** $0–100 a month ChatGPT or Claude for the team, a Zapier plan, a chatbot on the site. These help a person do a task faster. They do not take the task away, and nobody should sell you otherwise. — Every business, and where almost everyone should start before spending anything else. - **One workflow, wired up** $1,500–5,000 A freelancer or small shop connects two or three tools in Zapier, Make or n8n. When this fits, it genuinely fits. It stops fitting the first time something unusual arrives and there is nobody to catch it. — A single, well-defined, low-risk handoff between two systems. - **One process, end to end** $5,000–15,000 A system that reads the enquiry, document or booking, decides what to do under rules you set, acts across the tools you already run, and asks a person when it should. Integrated, staged, and tested on real data before it goes live. — Owner-led businesses where one repeatable process is eating a day a week of someone's time. - **Several processes, or one hard one** $15,000–50,000 The same thing where the hard parts live: approval queues, audit trails, a regulated environment, an in-house system with no API, or an operations dashboard your team steers all of it from. — Businesses with compliance obligations, or several processes that all touch the same data. - **Custom models and platforms** $50,000+ Training or fine-tuning on your own data, or building a product you intend to sell rather than a system you intend to run. — Product businesses. Very few of the companies asking this question need this. ### What actually moves the number. - **How many hours it takes back, and whose.** This is the first number to work out, not the last. It sets what the system is worth, which sets what it is sensible to spend. An assessment that does not open with this is selling you something. - **How many systems have to agree.** Two is a workflow. Five, one of which is a spreadsheet that somebody personally owns, is a project. The integration is rarely the hard part; agreeing which system is right when they disagree is. - **Whether a person has to approve anything.** An approval step means a queue, an interface to work it in, and a record of who said yes and when. It is the right call for anything customer-facing or money-related, and it is real work. - **How messy the source data is.** A clean API is a day. Emailed PDFs from fifteen suppliers, each laid out differently, half of them scans, is where the weeks go. - **Whether it has to stand up to an audit.** Regulated environments need change review, access control and a log of every decision the system made. We have run production systems for an ASX-listed financial services company since 2021, so this is a known quantity rather than a surprise. - **Volume.** Ten a week and a thousand a day are different systems, with different failure modes and a different monthly bill. ### Three real systems, line by line. Three of ours, cheapest first, so the band means something. This is the one at the bottom, and it shows what "attached to something we were already building" does to a price. Edtripper had two jobs. Every enquiry on the venue directory gets read as it arrives, judged real or spam, and forwarded to the attraction; when that attraction is not yet a partner the system looks their details up on their own site, checks what Edtripper holds is right, and builds the outreach itself. And inside the app, an attraction uploads its own education pack and the questions for a whole programme get drafted from it, multiple choice and photo challenges included, for an educator to sign off. Published with Edtripper's permission. Excludes GST. $5,000 is below our own automation floor of $8,000, for a reason worth stating. We were already building the platform these two automations run on: the auth, the data model, the deployment and the queues existed, and the second job was drafting against content the venue was already uploading. A standalone version of the same two jobs, integrated into somebody else's systems from the outside, is a $10,000 to $15,000 build. If you are already having something built, bolting the automation onto it is the cheapest it will ever be. - **Client** Founder-led education business, Sydney - **Scope** Enquiry triage and outreach, AI programme drafting - **Built in** 6 days - **Part of** a $39,000 app and dashboard build - **AI and automation** $5,000 6 days, on top of a platform we were already building for them **Costs to run** The automation itself costs nothing to keep online. The only variable line is model usage. - **Model usage** PAYG pay as you go, metered per venue - **Hosting and data** $0 Cloudflare Workers, D1, R2, Queues. Their own account - **Email** $0 Resend, on our business account at no charge to them ### The one most people actually buy. Joris is a buyer's agent in Newcastle whose business runs on admin. A property report for a client means generating it on a data service, pulling the details together, finding the client's record and writing the email. Enquiries arrive as Instagram messages, Facebook messages and website forms, and each one needs reading, judging and logging before anyone calls back. All of it was one person, after hours. Two systems. A custom MCP server he installed in the Claude and ChatGPT he already pays for, so "send Sarah the report on 12 Smith Street" is one sentence and the report is generated, matched to her HubSpot record and emailed. And a worker that reads every message and enquiry, qualifies it, files it in HubSpot at the right pipeline stage and writes the to-do with the context already in it. Published with Joris's permission. Excludes GST. - **Client** Buyer's agent, Newcastle - **Scope** Custom MCP server, property reports, lead qualification into HubSpot - **Built in** 4 weeks - **Live at** joris.com.au - **AI and automation** $15,000 4 weeks. Two systems, integrated with HubSpot and a property data service **Costs to run** No new logins and no per-task fees. He asks in the assistant he was already paying for. - **Model usage** PAYG pay as you go - **Hosting** $0 Cloudflare Workers, his own account - **Care plan** $350 covers these systems and the site we also built him - **New abilities** hourly scoped and billed by the hour as he asks for them ### What it costs to keep running. The bill that arrives every month after the build. It is smaller than most people expect, and the part everyone worries about, the model usage, is usually the smallest line on it. Excludes GST. Model pricing changes often; the usage figures above are what we see in production in September 2026. - **Model and API usage** $20–200 What the models actually charge for the work, for a typical SMB process. It is metered, so you can see it. Our own product runs a full launch pipeline, a deep crawl, venue discovery and every draft, for about a dollar of model spend per run. - **A workflow platform** $30–150 Zapier, Make or n8n, if the system uses one. Plenty of the systems we build do not, because the logic lives in code that you own rather than in someone else's canvas. - **Hosting and data** $0–50 Workers, a database and a queue. At the volumes an owner-led business runs at, this is usually inside a free tier. Of the three systems itemised on this page, two host for nothing and the third pays $10 a month for its database. - **Managed** $250–350 Monitoring, fixes when a model or an API changes underneath you, and a monthly report of what the system handled, what it saved and what it got wrong. Joris pays $350, which covers his automations and his website together, with new abilities scoped and billed by the hour as he asks for them. We require a managed plan for the first twelve months, because these systems drift and an unwatched one quietly stops being true. - **An agency support retainer** $500–2,000 What the Australian market charges for post-launch support and optimisation, per FlowWorks. Worth asking what it covers, and whether improvements are included or quoted. ### And the top of the range, ours. Nilkick is our own product, so there is no client invoice. We counted the hours anyway, at the rate we would have billed them, because a range with nothing at the top of it is not much of a range. It also answers the two questions the other receipts cannot. How long does a build like this really take: nine weeks, by the same three people who would build yours. And what does the model spend actually come to when a system is working hard: a full launch pipeline, a deep crawl, venue discovery and every draft, for about a dollar a run. No client invoice, so these are our own hours at the rate we would have charged. Excludes GST. Nine weeks of build at $36,000 is what the fourth band buys, and it is a product rather than an automation: five Workers, a payment system, an extension and a public API. Most businesses asking this question do not need anything like it. It is here because it is the only set of books we can open all the way. - **Client** Croissant & Baguette, own product - **Scope** Scanner, launch engine, Chrome extension, public API - **Built in** 14 days design, 9 weeks build - **Live at** nilkick.com - **Design** $6,000 14 days - **Development** $30,000 9 weeks. Five Workers, Workflows, cron, Supabase, Stripe, a Chrome extension and a public API - **Total** $36,000 hours counted at the rate a client would have been billed. Ex GST **Costs to run** Measured, not estimated. Every run meters its own model spend and timings. - **Model spend** PAYG about $1 per launch, metered on every run - **Database** $10 Supabase - **Hosting** $0 five Cloudflare Workers, Workflows and cron - **Search, payments and email** PAYG Serper and Bright Data pay as you go, Stripe per transaction, Resend on our own account ### What we charge. Indicative. Fixed price after the assessment. - **Automation Opportunity Assessment** $1,500 fixed (2 weeks · credited to your build) - **Automation build** from $8,000 (one process end to end, 30 days of hypercare) - **Managed** from $250/month (monitoring, fixes, monthly report · changes quoted) - **Internal tool / dashboard** typically $8–20k (on top of the systems doing the work) ### Ask away. Eight things people ask before they spend the money. Still one missing? Ask us. - **How much do AI automation agencies charge in Australia?** For one process automated end to end, $5,000 to $15,000 to build, and $500 to $2,000 a month if you take a support retainer. Published Australian tiers run $3,000 to $5,000 for one or two simple workflows, $5,000 to $15,000 for three to six, and $15,000 to $50,000 and up for full process automation. We start at $8,000, most builds land between $10,000 and $30,000, and our managed fee starts at $250 a month. - **How much does it cost to automate one process?** $5,000 to $15,000 for something defined, integrated with the systems you already run, and tested on real data before it goes live. What moves it is how many systems have to agree, whether a person approves anything, how messy the source data is, and volume. A single low-risk handoff between two tools costs less, from about $1,500, and it is a workflow rather than a system. - **What does it cost to run each month?** Usually $20 to $200 in model and API usage for an SMB process, plus $30 to $150 if the system uses a workflow platform, plus hosting that is often free at these volumes. Our managed fee is from $250 a month on top. Model spend is the line people worry about and it is almost always the smallest one: our own product runs a full launch pipeline for about a dollar. - **Is this just Zapier with extra steps?** Partly, and we use Zapier, Make and n8n where they fit. Workflow tools move data between apps when something triggers. What we build sits above that: it reads the enquiry, document or booking, decides what to do under rules you set, acts across your tools, and asks a person when it should. The workflow is one piece of it. - **How do we know it is worth it before we spend anything?** Work out the hours first. Count what the process takes each week, multiply by a loaded hourly cost of $65 to $75 for admin work or your own charge-out rate if the person is billable, and compare that to a build plus twelve months of managed fee. If the payback is longer than a year, do not do it yet. That arithmetic is what the $1,500 assessment does properly, across every process in the business, with build prices attached. - **What happens when the model or the API changes?** It breaks, eventually, which is why the managed plan exists and why we require it for the first twelve months. Models get deprecated, APIs change shape, and your own business changes what "correct" means. We watch it and fix it, and every month you get a report of what the system handled, what it saved and what it got wrong. - **Can we get funding for this?** Sometimes. Federal and state programs for technology adoption exist and change often. We will tell you what currently applies to your situation, and we will not build a business case that only works with a grant in it. - **What if we want to leave?** You own the code and the accounts. Everything is delivered into your repositories and your cloud, there are no per-task fees, and we hand over documented so it keeps running. That is the point of building it on infrastructure you own rather than renting you a seat on ours. - **Website costs.** What a website costs in Australia, by band, with a real receipt. - **Shopify store costs.** What a store costs to build, and what Shopify itself takes. - **AI Automation.** The five jobs we automate, and how the work is done. ### Book an assessment. Two weeks, fixed price. We map what is manual, what it costs you, and what to automate first, with prices attached. Credited against your build. ## Web Development https://croissantbaguette.com.au/web-development/ Sub-second websites, Shopify stores, web and mobile apps, built in Sydney and yours to keep. Websites from $9k, headless ecommerce from $15k, apps from $15k. **Service:** Custom website, ecommerce and web app development (Web development) Sub-second pre-rendered websites, headless and Shopify theme storefronts, and custom web apps, built in Sydney and owned by the client (code, CMS and hosting). - **Websites** from A$9000 — Pre-rendered marketing sites, editable by your team. From $9,000. - **Headless ecommerce** from A$15000 — Custom storefront on Shopify checkout. From $15,000, typically $18–35k. - **Shopify theme build** from A$9000 — Custom Shopify theme, launched in weeks. From $9,000. - **Web & mobile apps** from A$15000 — Portals, dashboards, booking systems, internal tools and mobile apps. From $15,000, typically $20–45k. - **Care plan** from A$250 — Monitoring, dependency updates, small changes and a monthly performance snapshot. From $250 a month. ### Sites, stores and apps that load in under a second and don't need us to update. The delivery layer for everything else we do. Custom-built, fast by default, and yours to keep: code, CMS and hosting. ### Three things we build. - **Websites.** Marketing sites, pre-rendered by default: sub-second loads, practically free hosting, nothing to patch, editable by your team. WordPress when it's the right call. - **eCommerce.** Shopify under the hood, a storefront that's actually yours. Fully headless or theme builds. - **Web & mobile apps.** Portals, dashboards, booking systems, internal tools, and the app your team carries. Built around how you work; AI-ready by default. - **+37% conversion rate** — Cheeky Kiwi Travel · booking platform, 10s to under 2s - **<0.5s page loads** — Joris · perfect performance scores - **30 days from kick-off to an Android app on live phones** — app, venue CMS and fleet cockpit in 22 staged releases - **33% conversion rate** — Made Quiet · headless Shopify, built in six days - **Double digit growth in sales** — Quitine · since moving off the theme ### Why builds still matter to us. Every AI system needs an interface, and every growth plan needs pages that load and get cited. Building is not what we lead with any more; it's what makes the rest real. sub-second, gets cited, yours to keep - **AI Automation** needs an interface — agent → UI - **SEO & AI Visibility** needs pages that load — SEO → pages **Web Development** ### Ranges. Fixed quote after a scoping call. No calculator; the interesting variables are never on a form. What moves the price within those ranges: - **Websites** from $9,000 - **Headless ecommerce** from $15,000 (typically $18–35k) - **Shopify theme build** from $9,000 - **Web & mobile apps** from $15,000 (typically $20–45k) - **Care plan** from $250/month - number of page templates and integrations - CMS and content-migration complexity - custom logic vs. configuration - how much design is from scratch ### Recently served. - Booking platform on Rezdy, 10s to under 2s, conversion rate up 37%. (https://croissantbaguette.com.au/projects/cheeky-kiwi-travel/) - ASX-listed financial services, 340 pages, published by their own team. (https://croissantbaguette.com.au/projects/fiducian/) - Headless Shopify store in six days. First sale in week one, 33% conversion rate. (https://croissantbaguette.com.au/projects/made-quiet/) - Free scanner to paid launch engine, extension and API, in nine weeks. (https://croissantbaguette.com.au/projects/nilkick/) - 27s to 0.7s. Same content, new build. (https://croissantbaguette.com.au/projects/wingboy/) - Headless Shopify store and the portal that pays 343 makers. (https://croissantbaguette.com.au/projects/pop/) ### Tell us what you need built. A site, a store, an app, or the thing that sits under all three. Fixed quote after one scoping call. ## Web & mobile apps https://croissantbaguette.com.au/web-development/web-mobile-apps/ Client portals, booking systems, dashboards and mobile apps built around how you work. iOS, Android and web from one codebase. From $15k, Sydney studio. **Service:** Web and mobile app development (Software development) Client portals, booking systems, dashboards, internal tools, marketplaces and mobile apps built around the client's workflow, hosted in the client's own cloud account and owned by them. - **Web app** from A$15000 — Portals, booking systems, dashboards and internal tools. From $15,000, typically $20–45k. - **Mobile app** from A$10000 — iOS and Android from one codebase, alongside a web app. From $10,000 on top, typically $10–30k. - **Care plan** from A$250 — Monitoring, updates and small changes from $250 a month, or managed at the same floor when AI is part of the system. ### Built around how you work, not the other way around. Client portals, booking systems, dashboards, internal tools, marketplaces, and the mobile app your team or your customers carry around. Custom logic, your data, no licence fees, and AI where it genuinely helps. **Spec ticket** Fixed quote after scoping. - **Runs on** your cloud account · Cloudflare, Supabase or Firebase - **Reaches** any browser · iOS · Android - **Typical build** 8–16 weeks - **After launch** care plan · or managed, when AI is in it - **+37% conversion rate** — Cheeky Kiwi Travel · booking platform, 10s to under 2s - **343 makers on one portal** — POP Local · sales, payouts and product submissions - **30 days first commit to live phones** — Edtripper · kiosk app, venue CMS and fleet cockpit - **9 weeks first commit to a paid product** — Nilkick · our own launch engine, five Workers ### Off-the-shelf only gets you so far. A few tools and workarounds work until they don't. Data ends up in five places, and the team spends its week moving it around. And when the work happens away from a desk (on a school excursion, in a warehouse, at a client's kitchen table), a browser tab isn't always the right tool either. At some point the cheapest option is the thing built for the way you actually work. ### Web apps, mobile apps, or both. Some businesses need one, some need both. We build either, and when you need both they share one backend and one team. - **Web app** the work happens at a desk or a phone browser is fine, you want it live fast, and nothing needs the device itself. — Cheeky Kiwi Travel (booking platform with an AI admin) · POP Local (maker portal and fortnightly payouts) · Edtripper (teacher directory and venue CMS) · Nilkick - Runs in any browser. No install, one codebase, updates the moment we ship them. - Logins, roles and permissions. Every user sees exactly what they should. - Integrated with what you already run: accounting, CRM, calendars, payments, booking tools. - Installable to the home screen as a progressive web app when that's all the “app” you need. - **Mobile app** the work happens off a desk, you need camera, location, offline or push notifications, or the devices belong to your organisation. — Edtripper: a kiosk-locked Android app on phones the venue owns, rolled out through MDM rather than the public stores. An update reaches every phone in under 90 seconds. - iOS and Android from one codebase. Alongside a web app, they share logic and backend; on its own, it gets its own. - Distributed through the App Store and Google Play, or privately to managed devices through your MDM. No public listing needed. - Works offline when the signal doesn't, and syncs when it does. - Push notifications, camera, location, and sign-in with the accounts your team already has. ### Web app and mobile app, compared. Plenty of businesses only need one: portals and back office live happily in a browser, field work wants the phone. When you need both, they share one backend, so the second is an addition, not a rebuild. - **Where it runs** Any browser, desktop or phone | On the phone or tablet itself - **Install and updates** None; updates are instant | From a store or your MDM; updates ship as releases - **Device features** Camera and location, with limits | Camera, GPS, push notifications, biometrics, background sync - **Offline** Limited | Built in - **Distribution** A link | Public stores, or private to devices your organisation manages - **Best for** Portals, dashboards, back office, bookings, anything done at a desk | Field work, frontline staff, devices your organisation owns ### Whichever way, you get this. - **Your data, your accounts.** Hosted in your cloud account, Australian region by default. We don't sit between you and your data. - **Roles and permissions.** Staff, clients, partners and admins each see their own slice, with an audit trail of who did what. - **Integrated, not isolated.** Xero, HubSpot, Google and Microsoft calendars, Stripe, Rezdy, and the in-house system nobody wants to touch. - **AI where it earns its place.** Natural-language search over your data, drafted replies, documents read and checked, with a human approving what matters. If it doesn't earn its place, we say so. - **Built for production.** Monitoring, error tracking, backups and a staging environment. Your team tests on real data before anything goes live. - **Yours to keep.** Code in your repositories, no licence fees, documented handover. It keeps running if we're not around. ### Apps we've shipped. - A booking platform on Rezdy's API, from 10-second loads to under 2 and a conversion rate up 37%, with an AI admin that answers booking questions in plain English. (https://croissantbaguette.com.au/projects/cheeky-kiwi-travel/) - A directory of 958 venues for teachers, and a kiosk-locked Android app for the venues, first commit to live phones in 30 days. (https://croissantbaguette.com.au/projects/edtripper/) - Trademark applications checked and processed around the clock. (https://croissantbaguette.com.au/projects/ttmf/) - Our own launch engine: free scanner, paid app, Chrome extension and public API, five Workers in nine weeks. (https://croissantbaguette.com.au/projects/nilkick/) ### Scope. Design. Build. Launch. - **Scope.** One or two workshops with the people who'll use it, a written spec, a fixed quote. If the point is removing manual work, this step is the assessment. - **Design.** Flows first, then screens. You click through a prototype before we write the real thing. - **Build in stages.** Weekly releases to a staging URL, your team on real data, the mobile app alongside if it's in scope. - **Launch and after.** Rollout, app-store or MDM release, monitoring, and 30 days of hypercare. Then a care plan, or the managed plan when AI is doing part of the work. ### Ranges. Fixed quote after scoping. No calculator; the interesting variables are never on a form. What moves the price within those ranges: - **Web app** from $15,000 - **Mobile app** from $10,000 (iOS and Android; on top of a web app) - **Care plan** from $250/month (or managed when AI is in it) - number of user roles and integrations - approval and workflow complexity - whether AI components need human-review steps - web only, or web plus iOS and Android ### Ask away. Six things people ask before they book. Still one missing? Ask us. - **Do you build mobile apps?** Yes. iOS and Android from one codebase, on their own or sharing a backend with your web app, released through the public stores or privately to managed devices through MDM. We'll also ask whether you need one: a web app installed to the home screen covers more cases than people expect. - **Web app, mobile app, or both?** Web app if the work happens at a desk or a phone browser is fine. Mobile if it happens in the field, needs the camera, location, offline or push, or runs on devices your organisation manages. Both when your office and your field team look at the same data; they share one backend. - **Can it connect to our existing systems?** Yes. Integration, not replacement, is the default: accounting, CRM, calendars, payments, booking tools, and whatever in-house system you already run. If it has an API we'll use it. If it doesn't, we've dealt with worse. - **Who owns it, and are there licence fees?** You own it: code in your repositories, hosting in your accounts. No per-seat fees, no per-task fees. Leave any time. - **How do we know it will keep working?** Monitoring and error tracking from day one, a staging environment for every change, backups, and 30 days of hypercare after launch. After that, a care plan for small changes or the managed plan when AI is part of the system. - **Is this the same as your AI Automation work?** They overlap. If you need software built around how you work, you're on the right page. If the point is to stop doing a process by hand, start with the assessment on the AI Automation page; most of those builds ship with an app like this on top. - **AI Automation.** Enquiries, documents and handoffs handled end to end, with a human signing off. - **Websites.** Pre-rendered and sub-second, or WordPress built properly. - **eCommerce.** Shopify under the hood, a storefront that's yours. ### Tell us what the team is doing by hand. Two lines is enough. You'll get a straight answer on whether it's an app, an automation or both, and a fixed quote after scoping. ## Websites https://croissantbaguette.com.au/web-development/websites/ Sub-second pre-rendered websites, or WordPress built properly. Editable by your team, yours to keep. 38× faster for Wingboy. From $9k, built in Sydney. **Service:** Website design and development (Web development) Marketing websites, pre-rendered on Cloudflare by default or built on WordPress when it suits the client, designed and built in Sydney and owned by the client (code, CMS and hosting). - **Website build** from A$9000 — Pre-rendered or WordPress, designed from scratch, editable by your team. From $9,000, typically $9–18k. - **Care plan** from A$250 — Monitoring, updates, small changes and a monthly performance snapshot. From $250 a month. ### Fast, cheap to run, yours to edit. Marketing sites for businesses that need them to work: load fast, rank, get cited by AI answers, and take a content change without a developer. Pre-rendered on Cloudflare by default. WordPress when it's the right call for you, built properly. **Spec ticket** Fixed quote after one call. - **Runs on** Pre-rendered · or WordPress - **Edited in** CloudCannon or Sanity · or the WordPress editor - **Typical build** 4–8 weeks - **After launch** care plan optional - **38× faster** — Wingboy · 27s → 0.7s - **1 → 4 people who can edit** — NIODA · 10s WordPress to 0.7s first paint - **3 → 1 three sites, one codebase** — Precise Group · WordPress - **<0.5s page loads** — Joris · perfect performance scores ### Most websites get slower, costlier and riskier every month. Plugins to update, hosting to pay, a database to protect, and a “quick change” that needs a developer. Wingboy's site took 27 seconds to load before we touched it. NIODA's took ten. That is rarely a platform problem; it is a neglect problem: a site nobody owns, on hosting nobody chose, running plugins nobody remembers installing. Slow sites don't just annoy. They don't rank, they don't get cited, and they don't convert. ### Two ways to build it. We'll recommend one after a 20-minute call. Both are designed and built by us in Sydney, and both are yours. - **Pre-rendered, fast** you want the fastest, cheapest-to-run site there is, and your site is mostly pages, forms and content. — Skyreach · Tilt Renewables · Wingboy · NIODA · Fiducian · Precise Group - Every page is built ahead of time and served from Cloudflare's edge. Sub-second loads, measured, not promised. - No server, no database, no plugins. Nothing to patch, nothing to hack. - Hosting that costs close to nothing. Often literally nothing. - Your team edits visually in CloudCannon or Sanity, with live preview, and publishes without us. - **WordPress, built properly** your team already knows WordPress, you rely on plugins with no real equivalent (memberships, courses, event ticketing, a directory). - A lean custom theme, not a page-builder stack. Fast on launch day and built to stay close to it. - Managed hosting, caching and a CDN set up from day one. - Only the plugins that earn their place, kept updated under a care plan. - The honest trade-off: it will never be as cheap to run or as hard to break as a pre-rendered site. We'll say so before you choose. ### Pre-rendered or WordPress? Most of the time we'll recommend pre-rendered. When WordPress is the right answer, we'll say so and build it well. - **Speed** Sub-second by design | Fast when hosted and cached well; slows as plugins pile up - **Hosting cost** Close to $0 a month | Roughly $30–100 a month for managed hosting - **Maintenance** None. No updates to run | Core, theme and plugin updates every month. The care plan covers it - **Security** No database, no login page to attack | Needs hardening, backups and updates. The most attacked platform on the web, because it's the most common - **Editing** Visual editor with live preview | The editor your team may already know - **Ecosystem** We build what you need; integrations through APIs | Tens of thousands of plugins. Memberships, courses and events out of the box - **Best for** Marketing sites, service businesses, anything where speed and cost to run matter | Content-heavy sites with features that live in the WordPress ecosystem, and teams that already run it ### Whichever way, you get this. - **Designed from scratch.** Your brand, not a template with your logo on it. - **Built to be found.** Semantic HTML, structured data, sub-second Core Web Vitals: the same foundation our Growth work relies on, and what AI answers cite. - **Editable by your team.** Visual editing with live preview. You're not waiting on us for a headline. - **Works on every screen.** Phones first, then everything else. - **Accessible by default.** WCAG AA contrast, keyboard navigation, real headings. Our own site scores 100 on Lighthouse accessibility, and yours will too. - **Yours to keep.** Code in your repository, hosting in your account, a CMS you own. Leave any time. ### Sites we've shipped. - A WordPress site rebuilt as a catalogue of 132 machines with real filters and a quote list. Five times faster. (https://croissantbaguette.com.au/projects/skyreach/) - 21 project mini-sites, a resource hub of 701 documents that builds itself, and twelve editors with the right permissions. (https://croissantbaguette.com.au/projects/tilt/) - 27 seconds to 0.7. Same content, new build. (https://croissantbaguette.com.au/projects/wingboy/) - A ten-second WordPress site one person could edit, rebuilt so four do. 391 changes published by NIODA staff since. (https://croissantbaguette.com.au/projects/nioda/) - Rebuilt in 2021, rebuilt again in 2026. 340 pages, 749 documents, and 99 of the last 118 edits made by their own team. (https://croissantbaguette.com.au/projects/fiducian/) - Three WordPress sites, one codebase. 309 pages, articles and projects ticked onto the sites they belong to. (https://croissantbaguette.com.au/projects/precise-group/) - A new brand, a site built to convert parents fast. (https://croissantbaguette.com.au/projects/rookies/) - A new agency in Newcastle, built to rank from day one. (https://croissantbaguette.com.au/projects/joris/) ### Scope. Design. Build. Launch. - **Scope.** A 20-minute call, then a fixed quote. We tell you which platform and why, in writing. - **Design.** Structure and copy first, then design. You review real pages, not mood boards. - **Build.** Built, connected to your CMS, loaded with your content, tested on real phones. - **Launch and after.** DNS, redirects, analytics, Search Console, and a month of watching the numbers. Then a care plan if you want one, or nothing at all: a pre-rendered site doesn't need us. ### Ranges. Fixed quote after a scoping call. No calculator; the interesting variables are never on a form. What moves the price within those ranges: - **Website build** from $9,000 (pre-rendered or WordPress) - **Content migration** quoted (from an existing site) - **Care plan** from $250/month (monitoring, updates, small changes, monthly performance snapshot) - number of page templates - CMS complexity and how much content moves across - integrations (forms, CRM, bookings, payments) - platform, because WordPress carries hosting and update overhead the care plan covers ### Ask away. Six things people ask before they book. The full numbers are in our website cost guide. Still one missing? Ask us. - **Do you build WordPress sites?** Yes, when it's the right fit: your team already runs it, or you need plugins that don't have a real equivalent. We build a lean custom theme on managed hosting and keep it updated under a care plan. If a pre-rendered site would serve you better, we'll say so first. - **Why is a pre-rendered site faster?** Every page is built ahead of time and served from a server near the visitor. Nothing runs when someone lands on the page: no database query, no plugin, no theme logic. That's why Wingboy went from 27 seconds to 0.7 with the same content. - **Can we update it ourselves?** Yes. Pre-rendered sites ship with CloudCannon or Sanity: click on the page, edit, preview, publish. WordPress sites use the editor your team may already know. - **What happens to our current rankings when we move?** We map every old URL to its new one with 301 redirects, keep the page titles that rank, and submit the new sitemap on launch day. Then we watch Search Console for a month. Done properly, a move doesn't cost you rankings. - **What does hosting cost?** Pre-rendered on Cloudflare: close to nothing, often literally nothing. WordPress: managed hosting at roughly $30–100 a month, in your own account. - **Do we need a care plan?** Pre-rendered: no. It runs without us; the care plan is for people who want changes made and a monthly performance snapshot. WordPress: yes, or someone on your side handles updates and backups every month. - **eCommerce.** Shopify under the hood, a storefront that's yours. - **Web & mobile apps.** Portals, bookings, dashboards, and the app your team carries. - **Growth.** Get cited when customers ask ChatGPT, Claude or Google's AI. ### Tell us what the site needs to do. Two lines is enough. You'll get a platform recommendation and a fixed quote within a business day of the call. ## Projects https://croissantbaguette.com.au/projects/ ### Skyreach https://croissantbaguette.com.au/projects/skyreach/ We rebuilt Skyreach's WordPress site on Astro and CloudCannon as a real catalogue: 132 machines with faceted filters, a quote list on every page, 5× faster. Client: Equipment hire company, eight branches on the East Coast · Industry: Equipment hire · Services: Static Website · Stack: Astro, CloudCannon, Cloudflare, Pagefind, Mapbox · Date: 2026-09 **Outcomes** - **5× faster** — fully loaded in 0.9 s, from 4.6 s, with a perfect performance score - **132 machines in the catalogue** — filtered by working height, capacity, outreach, width, weight, brand and power - **0 broken links** — 8,918 internal links checked; all 151 old product addresses still answer, no redirect - **129 spec sheets** — one per machine, on the page, with the specs in metric and imperial - **1 quote list on every page** — machines added from any card land in the enquiry form, already listed #### A catalogue that did not behave like one. Skyreach hires and sells boom lifts, scissor lifts, telehandlers and the rest of the height-access range from eight branches between Cairns and Melbourne. Thirty years in, with more than three thousand machines in the fleet, the website was a WordPress site on a page builder that had not kept up: slow, dated, and hard to move through. A machine was a page with a photo the size of a stamp, a paragraph, a form and a link to a PDF. Finding the right one meant knowing its name already. The data underneath was worse than the pages. The builder rendered two copies of every section, one for mobile and one for desktop, the product content lived only in rendered HTML, and most of the news existed as PDFs. #### A catalogue with real filters. A pre-rendered Astro site on Cloudflare, edited in CloudCannon. We pulled 132 machines, 129 spec sheets and 176 images out of the old site by script, extracted the specs into structured data and checked them, then gave every machine a proper page: specs in metric and imperial, features, the jobs it suits, the documents, the other models in its family. The catalogue filters by what matters on site: working height, lift capacity, horizontal outreach, machine width and weight, brand, power and features, with live counts and links you can share with a colleague. Search opens a dialog that only knows machines. Industry pages for mining, construction, events, government and telecoms put the right subset in front of the right visitor. Every old product address still works, without a redirect hop. #### A quote list instead of a form per machine. Hire enquiries rarely involve one machine. So every card and every product page has an add button, and the machines collect in a quote list that survives closing the tab and stays in sync across tabs. When the visitor is ready, the list travels with them into the quote form, every machine already named, hire or sale already marked. A machine's QR code opens the service-call form with that unit filled in, so a technician on site never types a fleet number. The branch map, the three forms and the resource centre are edited by the team. We handed over a written guide and a Claude skill that knows the site, so the team can ask their own assistant how to add a machine or an industry page. #### Same company, new site. The old homepage and product page next to the new ones, and the numbers behind them. The new site loads before the old one had painted its hero image. GTmetrix, September 2026, homepage, same location and device for both. - **GTmetrix grade** C → A - **Performance score** 62 → 100 - **Largest contentful paint** 3.3s → 559ms - **Fully loaded** 4.6s → 900ms #### The numbers. Catalogue and link counts from the build, September 2026. Speed from GTmetrix. #### Theirs to keep. The site lives in Skyreach's own CloudCannon organisation and Cloudflare account. The machines are markdown files with structured specs, so the catalogue can move anywhere that reads a folder. The import scripts and the extracted spec data are in the repository, with the site. - **Astro** the site and the catalogue, pre-rendered - **CloudCannon** editing machines, pages and the resource centre - **Cloudflare** hosting - **Pagefind** the machine search, rebuilt on each deploy - **Mapbox** the branch locator ### Tilt Renewables https://croissantbaguette.com.au/projects/tilt/ We rebuilt Tilt Renewables' WordPress site on Astro and CloudCannon: 21 project mini-sites and a 701-document resource hub that builds itself, 3× faster. Client: Renewable energy generator, Australia · Industry: Energy & infrastructure · Services: Static Website · Stack: Astro, CloudCannon, Cloudflare, Pagefind, Mapbox, Basin · Date: 2026-09 **Outcomes** - **3× faster** — fully loaded in 549 ms, from 1.6 s, with a perfect performance score - **21 project mini-sites** — each with its own pages, sidebar, documents and newsletter archive - **701 documents in one resource hub** — collected from 19 projects at every build, nothing to run by hand - **3 editing tiers** — twelve editors in four permission groups, publishing rights kept separate - **335 redirects** — every old WordPress address still lands somewhere #### Not slow. Just painful to change. Tilt Renewables develops and runs wind farms, solar farms and batteries across Australia. Each project is a small world of its own: planning documents, community pages, newsletters, benefit programmes, regulatory filings that must point at a stable address. The WordPress site had never been built for that. Project pages lived in a flexible-content field, sub-pages sat as flat top-level slugs, media was split across two hosts, and the projects were not even exposed to the API. Updating one project meant a developer, or a very patient editor. The site loaded in 1.6 seconds, so speed was not the complaint. Structure was. A team of twelve people needed to edit it, and not everyone should be able to touch everything. #### Every project is its own small site. A pre-rendered Astro site on Cloudflare, edited in CloudCannon. Each of the 21 projects is a folder: a project home with key facts, a location map, stages, stats, gallery and documents, then real sub-pages underneath it (planning and approvals, community, benefits), each with a stable address the regulators can cite. Every project gets its own sidebar, with a second level of anchors for the page you are on, and a newsletter switch that turns a sub-page into a dated newsletter with its own archive. The news archive, the 21 projects and 12 leader bios came across by script, with 335 redirects so nothing that was linked before goes dark. #### Who can edit what. Twelve people edit the site, in three tiers. Some can edit all of it. Some can only touch news and projects. Some can touch everything except the project pages. Four publishers, and only those four, can send changes live. All editing happens on a staging site; production is locked and only takes reviewed changes through a pull request. The permission groups run on CloudCannon's file-level rules, scoped by folder, so a project added next year inherits the same restrictions. We scripted the groups, so they can be rebuilt in a minute rather than clicked together again. #### A resource hub that builds itself. Every project carries its own documents: planning applications, approvals, newsletters, community reports. The Resource Hub gathers all 701 of them, across 19 projects and seven document types, into one filterable library with search and shareable links. Nobody maintains it. The hub is collected at build time from the project pages themselves, so it is right on every build, including the ones editors trigger from the CMS, and a check warns if a document is ever placed where the hub cannot see it. The same build indexes every page for search, writes the sitemap and llms.txt, and pushes every image through the optimiser, which editors cannot bypass by pasting a raw link. Documents and videos live in a Cloudflare R2 library the team uploads to from inside CloudCannon. #### The same pages, measured. The old site was not slow, so the gain is smaller than on most rebuilds. It is still three times faster, with a perfect performance score, on a site with far more in it. GTmetrix, September 2026, homepage, same location and device for both. - **GTmetrix grade** B → A - **Performance score** 87 → 100 - **Fully loaded** 1.6s → 549ms #### The numbers. Counts from the build, September 2026. The site keeps growing as the team publishes. #### Theirs to keep. The repository lives in Tilt's own GitHub organisation, the site in their own CloudCannon organisation and Cloudflare account, and every document in their own R2 bucket. The permission groups are a script they hold. If we disappeared tomorrow, the team would keep editing. - **Astro** the site, pre-rendered, 273 pages in about four minutes - **CloudCannon** visual editing, staging and publishing, four permission groups - **Cloudflare** hosting on Workers, plus the R2 document library - **Pagefind** search across every page, rebuilt on each deploy - **Mapbox** the location map on every project - **Basin** the four forms ### Edtripper https://croissantbaguette.com.au/projects/edtripper/ We built Edtripper's directory of 958 venues and a kiosk-locked Android learning app with venue CMS, fleet cockpit and AI-drafted questions, live in 30 days. Client: Founder-led education business, Sydney · Industry: Education & attractions · Services: Web Apps, AI & Workflows, Automation · Stack: SvelteKit, Sanity, Cloudflare, Android, Hexnode, Claude · Date: 2026-08 · Live site: https://edtripper.com **Outcomes** - **958 venues listed** — across every state and territory, September 2026 - **30 days first commit to live phones** — app, venue CMS and fleet cockpit in 22 staged releases - **2 days to a working browser demo** — then hardened on real hardware - **107 ms answer acknowledged** — p95 with 30 students answering at once, on the live realtime worker - **0 student names kept** — first name, initial and an avatar for the session, wiped at the end #### A product that never launched, and a site that had to do more. Edtripper began before COVID as a device-based learning product for museums and attractions. The week before its first venue went live, the venues closed. The business pivoted to a website for teachers, and the original app survived only as a sales deck, screenshots and memory. By 2026 the founder had two jobs for us. The website needed to become a real directory that teachers would use and the business could run without a developer. And attractions were asking for the device product again, so it had to be rebuilt from scratch: phones the venue owns, an app that survives museum wifi, and the teacher report that worksheets and guided tours both lack. #### A directory teachers actually use. A SvelteKit site, server-rendered on Cloudflare Workers, with every word in Sanity so the team publishes without a rebuild. Every visitor is resolved to their state (from the URL, then a cookie, then their location), and that state runs through every query: a teacher in Victoria sees Victorian venues, stories and lesson ideas first. The directory lists 958 venues across every state and territory, filtered by subject, year group and attraction type, with verified badges, amenities, booking links and curriculum outcomes on the articles that point to them. Structured data and per-type sitemaps on every page, because teachers find excursions through search. The founder runs content through their own AI assistant: we wrote a Claude skill that knows the content model, so they can draft, edit and publish venues and articles by asking for them. #### One app, three modes, on phones the venue owns. A single Android build, locked in kiosk mode on the venue's own fleet. Students get a trail of up to 30 stops with multiple choice, freeform and photo questions, a map of the floor, a leaderboard and a chat line to their teacher. The teacher gets a live view of every student, a broadcast channel and a check-in before the group leaves so no phone walks out. Venue staff get a guide view. Every live tour runs on its own realtime worker, and the app keeps working when the wifi does not. After the visit, every answer lands with the teacher as a signed web report and a class certificate. Students are a first name, an initial and an avatar for the session, and nothing more is kept. #### The venue CMS and the fleet cockpit. Venues write their own programmes: questions, zones, points, a map builder with floors and pins, a device preview, a test mode and immutable published versions so a live tour never changes under a class. Questions can be drafted by Claude from the venue's own education packs, but they land as drafts for an educator to sign off, never straight onto a device. The cockpit shows every phone's heartbeat, battery, kiosk state and app errors, logs incidents, and gives venue staff the safe actions (re-lock, reboot, rescan) with a pairing code to enrol a new phone. Device management is Hexnode behind an adapter of our own: when Google refused the device quota halfway through the build, the switch changed one file, not the platform. An app update reaches every phone in under 90 seconds. #### Every enquiry read, judged and answered. An enquiry on a venue listing used to be an email somebody had to deal with. Now it is read as it arrives: real enquiry or spam, and if it is real it goes straight to the attraction. When that attraction is not yet a partner, the system looks their details up on their own website, checks that what Edtripper holds is still correct, and builds the outreach itself, with the teacher's enquiry, an introduction to what Edtripper does, and an intro pack written for that venue. It is the same idea as the question drafting in the venue CMS, pointed the other way. The work that never gets done because nobody has time is exactly the work worth handing to a system, with a person still holding the final tap. #### The numbers. Build and load-test figures from August 2026. The first school group is the next number. #### Theirs to keep. Both products live in Edtripper's own Cloudflare and Sanity accounts, in their repositories, with the phones enrolled in their own device-management tenant. Nothing sits with us that they cannot take. - **SvelteKit** the website, the app and the venue CMS - **Sanity** content for the website, published by the team - **Cloudflare** Workers, D1, R2, Durable Objects and Queues - **Android** kiosk-locked app built with Capacitor - **Hexnode** device management, behind our own adapter - **Claude** question drafting and enquiry triage through AI Gateway, human-approved - **Resend** teacher reports, certificates and sign-in links ### Nilkick https://croissantbaguette.com.au/projects/nilkick/ We built Nilkick in nine weeks: a launch-readiness scanner, a paid launch app on Claude, a Chrome extension and a public API on five Cloudflare Workers. Client: Croissant & Baguette, own product · Industry: Founders & SaaS · Services: Web Apps, AI & Workflows · Stack: SvelteKit, Cloudflare, Supabase, Stripe, Claude · Date: 2026-07 · Live site: https://nilkick.com **Outcomes** - **9 weeks first commit to a paid product** — free report, three paid products, extension and API - **46 signals in the score** — 27 launch-readiness, 19 agent-readiness, none decided by an AI - **5 Workers in production** — the app plus four job workers on Workflows and cron - **≈$1 of AI per launch** — deep crawl, venue discovery and every draft, metered on each run - **402 automated tests** — plus a 22-scenario simulation of the warm-up plan #### Built isn't launched. Founders ship something good and it sits at zero. The advice they find is generic, and the real work (finding the venues that matter, reading the live threads where their users already talk, writing every post) is a grind nobody does well at eleven at night. We wanted a product of our own that did that grind with the founder keeping the final tap, and a test bed for the way we build AI systems for clients: metered spend, deterministic where it matters, a human in the loop by design. #### A score nobody can argue with. Paste a product URL. A Worker reads the page (with browser rendering as a fallback for JavaScript shells), checks how the domain is indexed and where it is mentioned, and scores it out of 100 from 27 rules across five groups: findable, clear, crawlable, indexed, referenced. A separate agent-readiness score runs 19 checks, from llms.txt to MCP and OAuth discovery. The integer is a pure, versioned function of what was detected. No model touches it, so the same site gets the same score on refresh and the report stays shareable at a stable URL. The launch map streams in behind it: real venues joined back from a curated pool of 100, so the model cannot invent one. #### Three products on one judgment engine. A Launch runs a deep crawl of the product, builds a brand and audience profile the founder approves, then produces a value-ranked queue of venues and live Reddit and Hacker News threads with every post already drafted, walked venue by venue. The Reddit warm-up is a 28-day plan of pure-value comments that builds account standing with no product mention, by construction. The Scout sweeps Reddit and Hacker News daily with the reply already written, and emails only on days with something new. The one rule: Nilkick never posts for you. Every submit is a human tap on the real button. Around it sits a marketing site with 57 guides, each with a markdown twin for AI crawlers, llms.txt and structured data from one content source. #### An extension and an API, for the people who want them. A Chrome side panel docks the same queue beside the venue tab: copy each field, tap submit yourself, and the done state writes back. Copy-only by decision: no scripts on third-party pages, no autofill. For anyone who wants to drive it programmatically, a public REST API with scoped keys, signed webhooks, batch claiming and rate limits that spend model money only where a draft is asked for. #### The numbers. Measured, not estimated. Every run meters its own model spend and timings. #### Ours, and the test bed for yours. Nilkick is where we try things before a client pays for them: model tiers picked per task, AI spend metered on every run, deterministic scoring next to model judgment. The same stack runs the systems we build for clients. - **SvelteKit** the app and the marketing site, with markdown twins for AI crawlers - **Cloudflare** five Workers, Workflows for durable jobs, cron, AI Gateway - **Supabase** Postgres and sign-in by email code - **Stripe** credits, the subscription and the customer portal - **Claude** Haiku for gates, Sonnet for drafting, Opus for deep ideation - **Serper · Bright Data · Algolia** search, Reddit and Hacker News data - **Resend** digests and transactional email - **Vite + CRXJS** the Chrome extension ### Quitine https://croissantbaguette.com.au/projects/quitine/ We moved Quitine from a Shopify theme to a headless SvelteKit and Sanity store, then gave the client a Claude skill and Sanity MCP to build pages by asking. Client: Pharmacist-founded nicotine replacement brand · Industry: Health & ecommerce · Services: Ecommerce, AI & Workflows · Stack: SvelteKit, Shopify, Sanity, Cloudflare, Claude, PostHog, Klaviyo · Date: 2026-07 · Live site: https://quitine.com **Outcomes** - **Double digit growth in sales** — since the move off the theme, by the client's count - **0 developers needed for a new landing page** — the client asks Claude; it drafts first and publishes on confirmation - **45 building blocks** — 10 atoms, 16 molecules, 6 wrappers and 13 sections, all editable by asking - **177 articles moved off Shopify** — into Sanity, with the old addresses redirected - **0 code changes to run an A/B test** — a variant page in Sanity and a PostHog flag, split on the server #### A theme can only do what a theme does. Quitine sells pharmacist-led nicotine gum, lozenges and patches, with a 90-day plan behind them. The store ran on a Shopify theme, and the theme had become the ceiling. The client wanted to put out a landing page in an afternoon, control every element on every page, and ship things a theme does not allow: a quiz that builds a kit and goes straight to checkout, chromeless pages for ad traffic, A/B tests on whole pages. The design came from the client's own team. Our job was everything underneath it: the store, the content system, and the way it gets edited. #### The store, headless. A SvelteKit storefront on Cloudflare Workers. Shopify keeps what it is good at: products, cart, checkout, customer accounts, subscriptions. Everything else moved to Sanity, including 177 articles brought across from the Shopify blog with their addresses redirected. Pages are built from a system of 45 blocks in four layers: atoms into molecules into wrappers into sections, each one a Sanity schema and a Svelte component. The same blocks make a product page, a resource article, or a listicle: a chromeless landing page for ad traffic with an offer or a quiz that lands in checkout. The Quit Quiz scores answers into nine outcomes with discount tiers; the kit builder prices a bundle live as it is assembled. A/B tests run on whole pages: a variant document in Sanity, a flag in PostHog, and the split decided on the server before the page is sent. #### Run by asking Claude. The client did not want a CMS to learn. He wanted to say what he needed. So the editing layer is a Claude skill we wrote: what the site is made of, every block and every field, how a page is put together, how the quiz works, how tracking and A/B tests are wired. Installed in his own Claude, next to the Sanity MCP and a small MCP of ours that uploads images, it turns a request into the work. New landing page for a campaign. Duplicate this page and change the hero. Add three FAQs. Import this article with its images. Set up a test on the homepage. The rules are in the skill: read before writing, drafts first, publish only when he confirms, never invent a field, ask before touching anything shared. It also says plainly when a request needs a developer, which is rare. When the content model changes, the skill is updated with it, so Claude never knows a version of the site that no longer exists. #### The numbers. The sales figure is the client's and is published as a range on purpose. Counts from the code, September 2026. #### Theirs to keep. Shopify, Sanity and Cloudflare are all Quitine's own accounts, and Claude runs on the client's own subscription. The skill is a folder in the repository, so the next person to work on the site inherits the same instructions the client uses. - **SvelteKit** the storefront, on Cloudflare Workers - **Shopify** products, cart, checkout, customer accounts, subscriptions - **Sanity** every page and article, with click-to-edit preview - **Claude** the editor, through our skill and the Sanity MCP - **PostHog** analytics and the A/B split - **Klaviyo** email, quiz results and cart recovery ### Joris https://croissantbaguette.com.au/projects/joris/ We built Joris a sub-second site and the AI systems behind it: an MCP his Claude and ChatGPT use for property reports, and a lead qualifier feeding HubSpot. Client: Buyer's agent, Newcastle · Industry: Property · Services: AI & Workflows, Static Website, Automation · Stack: SvelteKit, CloudCannon, Cloudflare, HubSpot, Claude, MCP · Date: 2026-07 · Live site: https://joris.com.au **Outcomes** - **<0.5s page loads** — perfect performance scores at launch, on a brand-new domain - **1 sentence to send a property report** — asked in his own Claude or ChatGPT, delivered by our MCP - **24/7 leads qualified** — every social message and website enquiry, into HubSpot with a to-do - **0 new logins for Joris** — everything runs through the assistant he already pays for #### New business, no reputation, and one person doing everything. Joris helps people buy houses in Newcastle. He started with no site, no rankings and established agents to compete with. The first job was to look established and be found. The second job showed up once the leads did. A buyer's agent runs on admin: a property report for a client means generating it on a data service, pulling the details together, finding the client's record, writing the email. Enquiries arrive as Instagram messages, Facebook messages and website forms, and each one needs reading, judging and logging before anyone calls back. Footage from inspections piles up on a phone, unposted. All of it was one person, after hours. #### A site that ranks on a new domain. A pre-rendered SvelteKit site on Cloudflare, edited in CloudCannon, that loads in under half a second with perfect performance scores. Service pages, success stories and resources built for the searches Newcastle buyers actually make, with structured data on every page. Google noticed, and so did the people searching for a buyer's agent. The enquiries that followed are what made the next two systems worth building. #### His assistant, our tools. Joris already pays for Claude and ChatGPT. Rather than give him another app to log into, we built a custom MCP server he installed in both, so he asks in a normal conversation and our tools do the work behind it. "Send Sarah the report on 12 Smith Street" is now one sentence. The MCP generates the report on the property data service, fetches every detail back, looks Sarah up in HubSpot, builds an HTML report and emails it to her, then tells him it is done. Because we host the MCP, adding a new ability is a matter of switching a tool on: no update for him to install, no new interface to learn. #### Every lead read, judged and filed. A worker receives every message from his social accounts and every submission from the website. Claude reads each one, qualifies it (a buyer with a budget and a suburb is not the same as an agent selling leads), creates or updates the contact in HubSpot, places it in the right sales pipeline stage and creates the to-do for Joris, with the context already written. He opens HubSpot to a list of people worth calling, not an inbox to triage. #### Footage in, posts out. The next system, in progress: Joris dumps the day's inspection footage into a folder. The AI watches it, picks the moments, cuts them into short edits, writes the captions and posts them to his social accounts on a schedule. Same rule as everything else we build for him: he approves the first ones, and the system earns its autonomy from there. #### The numbers. Lead and report counts are not published; the systems are measured on the managed plan's monthly report. #### Theirs to keep. The site, the MCP server and the lead worker run in Joris's own Cloudflare account, next to the HubSpot he already used. Every action the systems take is logged, and the assistants only ever see the tools we choose to expose. - **SvelteKit** the site, pre-rendered - **CloudCannon** editing the site without us - **Cloudflare Workers** the MCP server and the lead worker - **MCP** the protocol his Claude and ChatGPT use to reach our tools - **HubSpot** contacts, pipelines and to-dos - **Claude** reads and qualifies every lead, drafts the context ### POP Local https://croissantbaguette.com.au/projects/pop/ We rebuilt POP Local on headless Shopify, category pages from 20 seconds to under one, moved the shop off Square, and built the portal that pays 343 makers. Client: Independent retailer, Canberra · Industry: Retail & local makers · Services: Ecommerce, Web Apps, AI & Workflows · Stack: SvelteKit, Shopify, Sanity, Supabase, Cloudflare · Date: 2026-06 · Live site: https://poplocal.com.au **Outcomes** - **<1s page loads** — category pages took up to 20 seconds on the old build - **343 makers on the portal** — each with their own sales, payouts and product submissions - **1 system for shop and online** — Square and Next.js replaced by Shopify, point of sale included - **1 click fortnightly payout files** — bank files generated, reports emailed to every maker - **6k product URLs kept alive** — every old address redirected in the migration #### Twenty seconds a page, and the makers paid by hand. POP Local sells the work of hundreds of small Australian makers, in store and online. The old site, Next.js on top of Square, loaded every product and every category on every category page, so the pages people shop from took up to twenty seconds. Online sales were close to nothing. Behind the counter it was worse. Every fortnight someone built a sales report for each maker by hand and typed up a bank file for the payments. Makers dropped products at the shop with a name and a price, and the team wrote the descriptions with no idea what was in them. #### A store that loads, on Shopify. A headless storefront on SvelteKit reading products, collections, search and cart from Shopify's Storefront API, with the slow parts (recommendations, reviews, the maker's other products) streamed in after the page has painted. Category pages filter by maker, type, price and availability without a reload. Customer accounts sign in by email code or Google, with order history, loyalty points and reviews on top. Every page of content is in Sanity with a page builder and visual preview, so the team edits the site themselves. Click and collect runs through Shopify, and it is where online orders have taken off. #### Square to Shopify, counter included. The shop moved from Square to Shopify in one go: products, customers and the point of sale in store, so in-store and online sales sit in one system and every sale carries the maker who made it. The old catalogue had 1,208 product addresses; each one redirects to its new home, so nothing that ranked was lost. #### The portal that pays the makers. POP Maker is a web app for both sides of the counter. Makers apply, sign the agreement, then log in to see their own sales pulled live from Shopify, their payout history, and submit new products with the details the shop never had: ingredients, materials, care, allergens, the story. They keep their bank and GST details current themselves. Staff approve applications (which sets the maker up in Shopify, with a collection of their own on the site), review product submissions with an AI-enhanced description they can edit, and push products into Shopify as drafts. Every fortnight the payout run collects each maker's sales, works out the payout on cost of goods with GST added where the maker is registered, lets staff adjust and confirm, emails every maker their report, and generates the ABA files the bank accepts as a batch payment. What took a day of spreadsheets is a few clicks, with the numbers logged. #### The numbers. Maker count from the portal, June 2026. Online sales had no baseline to compare against; they started from close to zero. #### Theirs to keep. Shopify is the system of record for the shop, the site and the portal. Everything else runs in POP Local's own Cloudflare, Supabase and Sanity accounts, from their repositories. - **SvelteKit** the storefront and the maker portal - **Shopify** products, checkout, point of sale, sales data by maker - **Sanity** site content with visual preview, portal copy and the AI prompt - **Supabase** accounts, the portal database, payout files - **Cloudflare** hosting, plus a queue-driven worker for the payout run - **Growave** loyalty points, reviews and Instagram - **Resend · Klaviyo** maker emails and customer marketing - **Gemini** product descriptions drafted from what the maker submits ### Fiducian https://croissantbaguette.com.au/projects/fiducian/ We moved Fiducian off WordPress in 2021 and rebuilt it on Astro and CloudCannon in 2026: 340 pages, 749 documents, 99 of the last 118 edits by their staff. Client: Financial services group, ASX-listed · Industry: Financial services · Services: Static Website · Stack: Astro, CloudCannon, Netlify, Pagefind, Google Maps, SendGrid · Date: 2026-06 · Live site: https://www.fiducian.com.au **Outcomes** - **1.3s fully loaded** — performance score 97; the 2020 WordPress site took around 20 seconds - **340 pages** — 74 adviser profiles, 45 offices, 15 funds with 180 monthly reports - **749 documents** — product disclosures, reports and governance PDFs, checked on every build - **99 edits published by Fiducian staff** — in 90 days, out of 118 changes to the site; the other 19 were ours - **5 yrs without a developer for edits** — since the first rebuild in 2021; the second rebuild kept the CMS and the habit #### Twenty seconds to load, and locked out of their own site. Fiducian is a financial services group listed on the ASX: funds management, financial planning, a platform and software for advisers, and 74 planners across 45 offices. In 2021 the website was WordPress, took around twenty seconds to load, and the team could not change it. Every edit went through an outside agency, for a business that publishes fund reports monthly and announcements whenever the market requires. We moved them to a pre-rendered site with CloudCannon that year, and for five years their team ran it without us. In 2026 they came back for the next version: more advisers, more funds, more documents, a refreshed design, and a structure built for live editing from the ground up. #### Three hundred and forty pages, edited in place. A pre-rendered Astro site, edited visually in CloudCannon. Pages are assembled from 13 page sections and 27 building blocks, each with its own editor notes, so a new landing page is a matter of picking sections rather than asking us. Editors see the real page change as they type. The 102 articles, the 93 pages and the newsletter sign-ups are all theirs. The fund collection is deliberately locked to values only: fifteen funds, their sizes and 180 monthly reports can be updated by anyone on the team, but nobody can delete a fund by accident or break its structure. Every image goes through the build's optimiser, and search runs on a static index rebuilt on each deploy. #### Advisers, offices and documents as data. Seventy-four adviser profiles and 45 offices are linked records, not pages someone types out. An office lists its team by reference, so a planner who moves shows up in the right place the same day. The adviser finder uses Google Maps to rank offices by driving distance from a suburb, and the appointment form emails the planner, copies marketing and screens out bots without a captcha. Then the paperwork: 749 PDFs, from product disclosure statements to annual reports, moved by script and kept honest by a check that runs before every build and repairs any stray path. 366 redirects cover every old address, including the WordPress-era document links still cited elsewhere. #### 2020 and 2026. The WordPress homepage as it was before the first rebuild, and the site as it is now. The measured numbers are for the new site only; the old one is gone, and what Fiducian remembers of it is the twenty seconds. GTmetrix, September 2026, homepage. The WordPress site was replaced in 2021 and cannot be measured again. - **GTmetrix grade** A - **Performance score** 97 - **Largest contentful paint** 1.2s - **Fully loaded** 1.3s #### The numbers. Page, adviser and document counts from the build. Editing counts from the site's own change history, June to August 2026. Speed from GTmetrix, September 2026. #### Theirs to keep. The site lives in Fiducian's own CloudCannon organisation and hosting account. Advisers, offices, funds and pages are files in a repository their team commits to every week, so nothing is locked in a database or with an agency. - **Astro** the site, pre-rendered - **CloudCannon** visual editing across seven collections - **Netlify** hosting and the appointment endpoint - **Pagefind** search, rebuilt on every deploy - **Google Maps** the adviser finder, ranked by distance - **SendGrid** appointment emails to the planner ### Constellation Brands https://croissantbaguette.com.au/projects/constellation-brands/ We built the Shopify theme behind Constellation Brands' Fine Wine & Craft Spirits stores in AU and NZ: ten brands, member pricing, every section editable. Client: Wine and spirits producer, global · Industry: Wine & spirits · Services: Ecommerce · Stack: Shopify, Liquid, Theme editor · Date: 2026-05 · Live site: https://cbfinewine.com.au **Outcomes** - **AU + NZ stores on one theme** — the same sections and blocks on both storefronts, each with its own settings - **10 brands on one storefront** — from Robert Mondavi and The Prisoner to High West, each with its own page - **0 code changes to edit a page** — every section and block is set in the Shopify theme editor #### A global brand, and a theme that could not keep up. Constellation Brands is one of the world's largest producers of wine and spirits. Its Fine Wine & Craft Spirits arm sells Robert Mondavi, The Prisoner, Schrader, High West and the rest of the portfolio direct to customers in Australia and New Zealand. The theme behind those stores could not keep up with the brands, and the brief was bigger than one site: the same premium experience on both storefronts, cleared by a digital team in the United States, on the corporate-approved Shopify platform, to enterprise standards for policies, privacy and age verification. #### A custom theme, built for the theme editor. We designed and built the theme from a clean base, not a marketplace theme with a logo swapped in. Every section and block is defined for Shopify's theme editor, so the team rearranges pages, swaps brand showcases and changes copy without touching code, and both stores run the same theme with their own settings. Multi-brand showcases, a page per house, member pricing behind a sign-up, critic scores and awards on the bottle, and filtering by the things wine buyers filter by. #### Enterprise standards, met once. A company this size ships with paperwork: age verification on entry, responsible-drinking notices, privacy and personal-information controls, terms per market, and a review process run from another timezone. The build went through all of it once, so the Australian and New Zealand stores share the same approved base and the same policies, and a change to one is a change to both. #### The numbers. No traffic or sales figures are published for this client. #### Theirs, in Shopify. The theme lives in Constellation's own Shopify stores, versioned, with every setting in the theme editor. Nothing depends on us to change a page. - **Shopify** both storefronts, checkout and members - **Liquid** the theme, its sections and blocks - **Theme editor** every page, arranged by the team ### Made Quiet https://croissantbaguette.com.au/projects/made-quiet/ We built Made Quiet's headless Shopify store in six working days: first sale in week one, 33% conversion, sales just short of $100k in seven months. Client: New brand, first store · Industry: Apparel & ecommerce · Services: Ecommerce · Stack: SvelteKit, Shopify, Sanity, Cloudflare, Klaviyo · Date: 2026-02 · Live site: https://madequiet.co **Outcomes** - **6 days from first commit to live store** — after a week of design, January 2026 - **Week 1 first sale** — store opened February 2026, from a standing start - **33% conversion rate** — Shopify, August 2026 - **$100k just short of it in seven months** — sales February to August 2026, best month yet in August #### A brand about saying less needed a store that did the same. Made Quiet makes premium Australian sports caps for fans who do not need to shout. There was a name, a range and a clear identity, and no store. The brief was short: understated, fast, nothing in the way of the product, and live before the first drop in February. #### Headless Shopify, in six working days. A SvelteKit storefront on Shopify's Storefront API, deployed to Cloudflare Workers. A cart of our own that hands off to Shopify checkout, product pages that render first and stream the rest (recommendations, reviews) in behind, collection pages with filters and sorting, and structured data and sitemaps on every page from day one. Content sits in Sanity with a small page builder, so the team runs drops and campaigns themselves. Klaviyo captures the newsletter and back-in-stock requests; reviews come from Judge.me and show as star ratings on every card without an extra request. #### Nothing in the way of buying. The design is deliberately minimal: one typeface with character, the caps, and a path to checkout with nothing to dismiss. That restraint shows up in the numbers. The store sold in its first week, a third of visitors who arrive go on to buy, and seven months in the total is closing on six figures, with August the best month so far. #### The numbers. Sales and conversion figures from Shopify, shared with Made Quiet's permission and rounded on purpose. #### Theirs to keep. Shopify holds the products, orders and checkout. The storefront and content live in Made Quiet's own Cloudflare and Sanity accounts. Hosting costs close to nothing, and nothing needs us to update. - **SvelteKit** the storefront - **Shopify** products, cart and checkout through the Storefront API - **Sanity** pages and campaigns, edited by the team - **Cloudflare Workers** hosting at the edge - **Klaviyo** newsletter and back-in-stock - **Judge.me** reviews and star ratings ### Cheeky Kiwi Travel https://croissantbaguette.com.au/projects/cheeky-kiwi-travel/ We rebuilt Cheeky Kiwi Travel's WordPress site as a booking platform on the Rezdy API: from 10 seconds to under 2, conversion up 37%, 1.18 tours per booking. Client: Tour operator, New Zealand · Industry: Travel & tours · Services: Web Apps, AI & Workflows, Static Website · Stack: SvelteKit, Rezdy, Sanity, Cloudflare, Supabase, Claude · Date: 2025-12 · Live site: https://cheekykiwitravel.com **Outcomes** - **+37% conversion rate** — sessions that book, 1.83% to 2.51% - **+26% conversion rate, per visitor** — 2.33% to 2.94% - **+15% cross-sell** — 1.18 tours per booking, up from 1.03, the second one sold in the cart - **<2s page loads** — down from 10 seconds or more on the old site - **2 brands, one codebase** — Cheeky Kiwi and Crew Adventures share the platform and the CMS #### Ten seconds to load, on the page people book from. Cheeky Kiwi runs some of New Zealand's most popular small-group tours. The site did not match: WordPress with the Rezdy booking plugin, ten seconds or more to load a tour page, a dated design, and visitors bouncing before the calendar appeared. Every booking still had to go through Rezdy, so the answer was not another theme. It was a site that talks to Rezdy properly. Behind the scenes, every afternoon someone spent one to two hours turning the day's bookings into headcounts and emails for more than fifteen suppliers, by hand. #### A booking platform on the Rezdy API. A custom platform on SvelteKit. Tour and editorial pages are pre-rendered from Sanity, so they load before Rezdy is even asked. Availability and pricing stream in from the Rezdy API behind them, two months at a time. A cart of our own holds several tours and departures at once, quotes live, takes vouchers and coupons, and offers the rest of a package when part of it is in the cart. That is where the second tour gets sold. Cards are tokenised in the browser with Rezdy's own Stripe key, so no card detail ever touches the site. Rezdy stays the source of truth: a webhook keeps every tour in sync with the CMS, and the team edits everything else in Sanity. A second brand, Crew Adventures, runs from the same codebase and dataset. #### The 3pm supplier run, drafted by an AI admin. Bookings mirror from Rezdy into a staff tool as they land. Each afternoon it groups tomorrow's passengers by supplier, applies each supplier's rules (cut-off times, capacities, seasonal allocations) and has Claude draft every supplier email. Staff read, adjust and send, and every email is logged. A one-to-two-hour job becomes a review. The same tool answers booking questions in plain English: how many passengers tomorrow, which bookings carry dietary notes, what extras are on Thursday. It can only answer from the live data it fetches, and it does its arithmetic with a calculator, not by guessing. After the trip, guests get one question. Five stars sends them straight to TripAdvisor; anything less goes to a private form first. The latest reviews sync back to the site every day. #### The numbers. Conversion and cart figures compare the old site with the new one, published with Cheeky Kiwi's permission. #### Theirs to keep. The platform lives in Cheeky Kiwi's own accounts, integrated with the Rezdy they already run. Bookings, payments and availability stay where they were; only the site in front of them, and the admin behind them, changed. - **SvelteKit** the site, the cart and the staff tool - **Rezdy** tours, availability, bookings and payments, through the API and webhooks - **Sanity** content for both brands, with a rebuild button in the studio - **Cloudflare** hosting, plus workers for the Rezdy, TripAdvisor and rebuild hooks - **Supabase** the booking mirror and staff logins - **Claude** supplier emails and the booking chat, tool-bound to live data ### Rookies https://croissantbaguette.com.au/projects/rookies/ We built Rookies Club a bold, conversion-focused SvelteKit site to launch their teacher-led, athlete-backed kids sports programs across Australia. Services: Static Website · Stack: SvelteKit, Cloudcannon · Date: 2025-10 · Live site: https://rookiesclub.com.au #### A new brand needed a site that hits as hard as its mission. Rookies Club is a fresh take on kids' sports—teacher-led programs backed by elite athletes, run right at school. They needed a website that could sell that vision and convert parents fast. We designed and built a high-energy, fully pre-rendered site from scratch. Every page loads instantly, the brand punches through, and the registration flow is frictionless. The site runs on Cloudcannon with full live preview, so the team can update programs, FAQs, and content without touching code. Built to scale as Rookies rolls out across Australia. ### Precise Group https://croissantbaguette.com.au/projects/precise-group/ We rebuilt Precise Air, Precise Fire and AirGene as one SvelteKit codebase and one CMS: 309 pages across three sites, 422 redirects, publishing without us. Client: Building services group, national · Industry: Mechanical, fire & electrical services · Services: Static Website · Stack: SvelteKit, CloudCannon, Cloudflare, Pagefind, Basin · Date: 2025-09 · Live site: https://www.preciseair.com.au **Outcomes** - **3 → 1 sites, one codebase** — Precise Air, Precise Fire and AirGene from one repository, one CMS, three builds - **309 pages, articles and projects** — each ticked onto one, two or all three sites from a single field - **273 edits by the Precise team** — published by two people since launch, more than the ones we made - **831ms largest paint** — GTmetrix grade A on the Precise Air homepage - **422 redirects** — every article on the three old sites still lands somewhere #### Three websites, three times the work. Precise Group is a national building services company: Precise Air for mechanical services, Precise Fire for fire protection, AirGene for air hygiene, with offices from Perth to Brisbane. Each company had its own WordPress site. All three were slow, all three looked like different decades, and a piece of news relevant to the group had to be posted three times, in three admin panels, by whoever had the login. The brands are distinct on purpose. The work behind them is not. The brief was one system that could tell them apart. #### One codebase, three sites, one tick box. A single pre-rendered SvelteKit codebase that builds three times, once per brand, from one content folder. Every page, capability, article and project has a site field: tick Precise Air, tick Precise Fire, tick AirGene, or all three, and it appears on each with that brand's navigation, footer, colours and metadata. 309 content items today, 173 of them news, and the ones about the group are shared. Shared components, three accents: the navy and red of Precise Air, Fire's orange, AirGene's green. The team edits it all in CloudCannon with live preview. A story about the group is written once and published to whichever brands it belongs to. #### Handed over, and used. The leaders, 61 projects, 42 capability pages and the news archive came across, with 422 redirects so that nothing linked from the old three sites goes dark. Since then two people at Precise have published 273 changes themselves: new projects, news, careers, the safety and ESG pages. Images live in a Cloudflare bucket and are resized on the edge, search runs on a static index per brand, and enquiry forms post to Basin. #### Same company, new sites. The old Precise Air homepage next to the new one. The numbers are for the new site; the three old ones are gone. GTmetrix, September 2026, Precise Air homepage. The WordPress sites cannot be measured again. - **GTmetrix grade** A - **Largest contentful paint** 831ms - **Fully loaded** 1.3s #### The numbers. Content and redirect counts from the build. Editing counts from the site's change history, May 2025 to July 2026. Speed from GTmetrix, September 2026. #### Theirs to keep. The three sites build from one repository in Precise Group's own CloudCannon organisation, with images in their own Cloudflare bucket. Adding a fourth brand is a fourth build and a fourth tick box, not a fourth website. - **SvelteKit** three sites from one codebase, pre-rendered - **CloudCannon** one CMS, a site field on every item - **Cloudflare** hosting, the image bucket and edge resizing - **Pagefind** search, one index per brand - **Basin** the enquiry forms ### NIODA https://croissantbaguette.com.au/projects/nioda/ We rebuilt NIODA's ten-second WordPress site on SvelteKit and CloudCannon: 221 pages and records, and four staff publishing 391 changes without us. Client: Postgraduate institute, Melbourne · Industry: Education · Services: Static Website · Stack: SvelteKit, CloudCannon, Cloudflare, Pagefind · Date: 2025-09 · Live site: https://nioda.org.au **Outcomes** - **1 → 4 people who can edit the site** — 391 changes published by NIODA staff since launch, none through us - **0.8s fully loaded** — the WordPress site took ten seconds or more a page - **221 pages and records** — 78 insights, 58 events, 45 faculty and staff profiles, 36 pages, 4 theses - **40 page sections** — courses, events, programmes and profiles built from the same editable blocks - **133 redirects** — every old WordPress address still lands somewhere #### Ten seconds a page, and one person who could change it. NIODA, the National Institute of Organisation Dynamics Australia, teaches leaders and consultants how organisations really work: a master's degree, a graduate diploma and certificate, a PhD, workshops, events and a growing body of research. The website was WordPress on a drag-and-drop page builder. Every page had been assembled by hand, so there was no template to follow, and every page took ten seconds or more to load. Only one person could safely change it, and that person was the CEO. A new event, a staff bio, a symposium programme: all of it queued behind one desk. #### A structure the whole team can edit. A pre-rendered SvelteKit site on Cloudflare, edited in CloudCannon. Instead of hand-built pages, six collections with a schema each: pages, faculty and staff, insights, events, theses and site data, with 40 page sections the team assembles visually and sees change as they type. Courses share the same details, pricing and what-to-expect blocks, so a new workshop looks like the rest without anyone designing it. Staff profiles are proper records: qualifications, teaching, consulting practice, research and publications, with every insight linked to its author. Events carry Melbourne-time sessions, programmes and their own FAQs, and sort themselves into upcoming and past. #### One editor became four. NIODA staff were editing in CloudCannon less than two weeks into the build, well before launch, so the handover was done by the time the site went live. Since then four members of the team have published 391 changes themselves, and the CEO is no longer the bottleneck. Our own edits in the last six months: four. Images live in a Cloudflare bucket and are resized on the edge, search runs on a static index, and 133 redirects keep every old WordPress address alive. #### Same institute, new site. The old page-builder homepage next to the new one. The numbers are for the new site; the old one is gone, and its ten seconds is what it took when we first looked at it. GTmetrix, September 2026, homepage. The WordPress site cannot be measured again. - **GTmetrix grade** A - **Performance score** 98 - **First paint** 656ms - **Fully loaded** 788ms #### The numbers. Content counts from the build. Editing counts from the site's change history, August 2025 to May 2026. Speed from GTmetrix, September 2026. #### Theirs to keep. The site lives in NIODA's own CloudCannon organisation and Cloudflare account. Every course, event, profile and insight is a file in their repository, so the content moves anywhere that reads a folder. - **SvelteKit** the site, pre-rendered - **CloudCannon** six collections, edited visually by four staff - **Cloudflare** hosting, the asset bucket and edge image resizing - **Pagefind** search, rebuilt on every deploy - **JotForm** enrolment and contact forms, embedded ### Growing Spaces https://croissantbaguette.com.au/projects/growing-spaces/ We merged Growing Spaces' WordPress and Shopify setup into one SvelteKit and Shopify site: articles, products and pages in one place, run by their team. Services: Ecommerce · Stack: SvelteKit, Sanity, Shopify · Date: 2025-02 · Live site: https://growingspaces.com #### Two websites became one that actually works. Growing Spaces sells award-winning dome greenhouses across the US. Their online setup was less elegant — WordPress for content, Shopify for sales, and confused customers caught in between. We merged everything into a single site with one database and one experience. Articles, products, and pages all live together now. The team manages it all from one place, and customers can browse and buy without the runaround. Cleaner for everyone. ### Creative Resort https://croissantbaguette.com.au/projects/creative-resort/ We built Creative Resort, a Sydney branding studio, a SvelteKit and CloudCannon site to replace Squarespace: work shown properly, easy to update. Services: Static Website · Stack: SvelteKit, Cloudcannon · Date: 2024-12 · Live site: https://creativeresort.com.au #### A rebrand deserved a site to match. Creative Resort is a Sydney branding studio. They'd just finished rebranding themselves, but their old Squarespace site wasn't keeping up. We built them something custom — flexible enough to show off their work properly, and simple enough that they can update it without fighting the platform. No more clunky workarounds or limited options. Their brand looks the part now. So does their site. ### 1300 Australia https://croissantbaguette.com.au/projects/1300-australia/ We redesigned 1300 Australia's homepage and enquiry flow so more visitors become leads, with building blocks their team uses to update the site themselves. Services: Static Website, Web Apps · Stack: SvelteKit, Cloudcannon · Date: 2024-10 · Live site: https://1300australia.com.au #### Turning phone number shoppers into actual enquiries. 1300 Australia sells memorable business numbers — the kind with words in them. Their site needed to do a better job of closing the deal. We redesigned the homepage and rebuilt the enquiry form to make the whole process smoother. We also set them up with building blocks so they can maintain and update the site themselves. They pay less. Visitors convert more. The maths works out. ### Kea Health https://croissantbaguette.com.au/projects/kea-health/ We rebuilt Kea Health's WooCommerce store, which buckled under traffic, as a headless SvelteKit, Shopify and Sanity site: fast, reliable, ready for spikes. Services: Ecommerce · Stack: SvelteKit, Shopify, Sanity · Date: 2024-10 · Live site: https://keahealth.com #### A store that couldn't keep up with demand. Kea Health sells products that help people quit smoking. Their old WooCommerce site kept buckling under traffic — slow loads, lost sales, frustrated customers. We rebuilt the whole thing. Svelte on the frontend, Shopify handling the backend, and Sanity for content. The site's fast, reliable, and ready for traffic spikes. More people make it to checkout now. More of them leave as ex-smokers. ### Karis Life https://croissantbaguette.com.au/projects/karis-life/ We rebuilt Karis Life's crashing WooCommerce store as a headless SvelteKit, Shopify and Sanity site: fast, stays up under load, edited without a developer. Services: Ecommerce · Stack: SvelteKit, Shopify, Sanity · Date: 2024-09 · Live site: https://karislife.com.au #### Their store kept crashing. Their customers needed it not to. Karis Life sells assistive technology to people who depend on it. Their WooCommerce site couldn't handle the growth — crashes, downtime, and lost orders. We rebuilt it with Svelte, Shopify, and Sanity. The site loads fast, stays up, and the team can update content without calling a developer. No more crashes. No more missed sales. Just a store that works when people need it. ### TTMF https://croissantbaguette.com.au/projects/ttmf/ We built TTMF a trademark registration app: upload a logo, pay, and AI checks its chances in one flow. SvelteKit, Firebase, Stripe and OpenAI, running 24/7. Services: Web Apps, AI & Workflows · Stack: SvelteKit, Firebase, Cloudcannon, Stripe, OpenAI · Date: 2024-08 · Live site: https://ttmf.com.au #### Trademark registration without the back-and-forth. TTMF helps people register trademarks. The traditional process involves lawyers, paperwork, and waiting. We built something faster. Upload your logo, pay the fee, and AI checks if it's likely to work — all in one flow. Svelte frontend, Firebase backend, Stripe for payments, and OpenAI doing the heavy lifting. It runs 24/7, never takes a day off, and gets people through the process without the usual friction. ### Wingboy https://croissantbaguette.com.au/projects/wingboy/ We rebuilt Wingboy's 27-second WordPress site on SvelteKit and CloudCannon: it now loads in 0.7 seconds, 38 times faster, across three Sydney wing shops. Services: Static Website · Stack: SvelteKit, Cloudcannon · Date: 2024-06 · Live site: https://wingboy.com.au #### From 27 seconds to instant. Wings ordered before hunger kicks in. Wingboy sells wings across three Sydney locations. Their WordPress site took 27 seconds to load — long enough to lose customers to the competition. We rebuilt it. The new site loads in 0.7 seconds. That's 38 times faster. People can find the menu, pick a location, and order before the craving fades. Fast site, more wings sold. ### Beige Tech https://croissantbaguette.com.au/projects/beigetech/ We rebuilt Beige Tech's site on SvelteKit and CloudCannon with redrawn animal illustrations, scroll animations and a visual CMS their team edits themselves. Services: Static Website · Stack: SvelteKit, Cloudcannon · Date: 2024-05 · Live site: https://beige.tech #### A software company that wanted a site with personality. Beige builds tools for energy companies. Their old site was fast and scored well, but it was locked down and looked forgettable. We rebuilt it with a visual CMS so their team can add pages and update content without touching code. We also redrew their signature animal illustrations and added animations — some built in After Effects, some pure CSS — that respond to scrolling and actually serve a purpose. Now it looks like them, moves like them, and they don't need us to keep it running. ### Beamible https://croissantbaguette.com.au/projects/beamible/ We built Beamible a fast Hugo and CloudCannon site that doesn't look like every SaaS template: their marketing team ships landing pages without a developer. Services: Static Website · Stack: Hugo, Cloudcannon · Date: 2024-03 · Live site: https://beamible.com #### A SaaS site that doesn't look like every other SaaS site. Beamible helps businesses design better ways of working. They wanted a website that felt as thoughtful as their product — fast, clean, and not boring. We built them a site that loads quickly and stands out from the usual SaaS template look. Their marketing team now spins up landing pages on their own, updates content when they need to, and never has to wait on a developer. That's how it should work. ### Youxia https://croissantbaguette.com.au/projects/youxia/ We replaced Youxia's slow WordPress site with a pre-rendered SvelteKit and CloudCannon site that loads instantly and is updated without a developer. Services: Static Website · Stack: SvelteKit, Cloudcannon · Date: 2020-05 · Live site: https://youxiacrypto.com #### WordPress was the problem. We removed it. Youxia helps people buy crypto and gold. Their old WordPress site was slow and frustrating to manage. They wanted something better. We built them a static site that loads instantly. Pre-rendered pages, no bloat, and a CMS that lets them update content without touching code or waiting on developers. Problem solved. Speed delivered. ### Thermotec https://croissantbaguette.com.au/projects/thermotec/ We built Thermotec, an insulation maker of 35 years, a headless store on SvelteKit, Shopify and Sanity: fast, reliable, and updated by their own team. Services: Ecommerce · Stack: SvelteKit, Shopify, Sanity · Date: 2020-05 · Live site: https://thermotec.com.au #### 35 years of insulation. A website that finally keeps up. Thermotec has been making insulation for over three decades — keeping things warm, keeping things quiet. They needed an online store that worked as reliably as their products. We built it with Svelte on the frontend and Shopify handling orders. Sanity gives them full control over content, so they can update pages and add products without needing us. Fast, independent, and built to last. ### Pet Professor https://croissantbaguette.com.au/projects/pet-professor/ We built Pet Professor, a dog training school, a fast Hugo and CloudCannon site that ranks for local searches and makes booking a class simple. Services: Static Website · Stack: Hugo, Cloudcannon · Date: 2020-05 · Live site: https://petprofessor.com.au #### Helping dogs behave and owners find help. Pet Professor runs dog training classes. They needed a site that looked good, loaded fast, and showed up when people searched for help with their unruly pup. We built them something clean and easy to navigate. It ranks well on Google, loads quickly, and makes booking classes simple. Dogs learn to sit. Owners find the site. Everyone's happy.