When should an agency say no to a site builder and build custom?

By Ranjit Bhinge, founder of Blur Studio · Updated · 9 min read

Short answer: Say no to a site builder when the client needs something the builder cannot do or cannot let go of: logins and accounts, custom logic, deep connections to other systems, strict data control, or a site they must be able to take with them. A site builder is a hosted tool such as Wix, Webflow or Framer that lets you assemble pages visually. For a marketing site with standard pages, a builder is usually cheaper and faster. Go custom when the cost of working around the builder over three years passes the cost of building it properly once.

Ranjit Bhinge

A client has asked for something a site builder only half does, and you need to decide whether to stretch the tool or build it properly.

I'm Ranjit, and I run Blur Studio: 15+ years building software, 120 apps taken live, and clients who have gone on to raise $10M+. I build AI systems and processes that help businesses do a lot more with a lot less time. I also work differently from most of what else is out there, more on that further down this page. Say hello.

The sensible default is to use a builder. For a brochure site, a small shop or a landing page, a hosted builder gets the client live sooner and costs less. Clutch, a review site for agencies, says about 75 percent of small business websites are powered by builders (it gives no sample size, so treat it as a rough guide). The question is not "are builders bad". It is "has this job outgrown one". If you can answer yes to two or more of the seven signs below, say no to the builder and quote custom, or at least quote a short test first.

This page is written for agency owners, but a business owner can use it to judge what an agency proposes. Prices are in USD as of November 2026 and come from the vendors' pages or the sources listed. Check the vendor page before you quote, because plans change. Hostinger is a web hosting company, WebsiteSetup is a website guide site, and both publish cost guides.

Seven signs you need custom code or an app

  1. Users log in and see their own data. A client portal, member area, booking history or dashboard needs accounts, permissions and a database. A builder may offer a membership add-on, but you are then limited to its rules. If the data is the product, build it.
  2. The business rules are the point. Quotes that depend on ten inputs, matching jobs to technicians, or stock that changes by branch are logic, not pages. If you find yourself writing workarounds in embedded code snippets, the logic wants a proper home.
  3. Several systems must stay in step. If the site must read and write to the client's CRM (customer records tool), accounting, scheduling and inventory, and a connector tool such as Zapier (a service that links apps) keeps breaking or costs more than the build would, custom wins. Native connections on a builder are limited to what the vendor chose to offer.
  4. Data rules are strict. If the client handles health, financial or children's data, or must keep data in one country, you need control over where it lives and who can reach it. A hosted builder decides that for you. Check local law and ask the platform in writing.
  5. The client must be able to leave. Export rules differ. Per the vendors' own pages as of November 2026: Wix says the site must run on its servers, Framer does not offer HTML export, Squarespace exports limited content to an XML file for WordPress, a widely used free website system, and Webflow lets you export HTML, CSS and JavaScript on paid Workspace plans, but its pricing page says forms stop working and dynamic content (CMS items) must be exported collection by collection. Lovable, an AI tool that builds apps and sites from a written brief, says on its documentation that you can export your code. If handover and exit are contract terms (see contracts for website work in the AI era), a platform with no export is a poor fit.
  6. Speed or scale is a hard requirement. A site with thousands of pages, heavy filtering or live data may hit page, item or bandwidth limits. For example, Webflow's Team plan page lists limits such as 500 static pages and 20,000 CMS items. Check the limits for the plan you would buy, not the headline.
  7. The platform's price or policy is a risk you cannot carry. Hosted builders can reprice or change plans, and Webflow changed its plan prices in 2026 (check its pricing page). If the client's margin cannot absorb a price rise, or you cannot pass it on, you carry the risk of someone else's pricing.

None of these means a builder is wrong for every site. It means this job has a specific need the tool does not meet well.

When the builder is the right answer

Do not go custom just because you can. A builder is the better pick when:

  • The site is mostly pages, a blog, a contact form and a booking link.
  • The client needs to edit text and images themselves without calling you.
  • The budget is small and the site must be live in days, not months.
  • You can name the platform's limits and none of them touch this client.
  • The client is happy to run on a monthly subscription and accepts that the site lives on the platform.

A related rule: AI tools now build decent pages in an hour, which makes the builder case even stronger for simple sites. See what an AI website builder is and the builders agencies compare. The harder case is the one in the middle, which is where this page helps.

The three-year cost view

Compare the two routes over three years, not at launch. Launch price is where builders look best. The gap closes when you add the monthly cost, the workarounds and the cost of leaving.

Real reference points, from Clutch and WebsiteSetup as of 2026 (both are third-party roundups, and sellers' numbers vary a lot, so use them as ranges and not as quotes): builder subscriptions run from about $17 to $159 a month for Wix, $16 to $25 for Squarespace and $3 to $30 for WordPress hosting plans. Agencies bill about $25 to $150 an hour on Clutch's range, and WordPress developers $50 to $95 an hour on WebsiteSetup's. A basic 5 to 10 page site from a freelancer is quoted at $500 to $5,000, and a custom build commonly lands at $5,000 to $30,000 or more on Hostinger's range.

Here is a made-up example to show the method. Every number is invented.

Item (made-up)Builder routeCustom route
Build, one time$4,000$18,000
Platform or hosting, per month$60$40
Paid add-ons and connectors, per month$90$10
Workarounds and fixes, hours per month at $956 hours, $5701 hour, $95
Care and updates, per month$0 (included in the above)$150
Monthly running cost$720$295
Three-year running cost (36 months)$25,920$10,620
Three-year total$29,920$28,620
Cost to leave, one time$6,000 (rebuild elsewhere)$2,000 (move the code and data)
Total including exit$35,920$30,620

In this invented case the builder looks $14,000 cheaper on day one and costs more by month 36 once the workarounds and the exit are counted. Change the workaround hours to 1 a month and the builder wins easily. That is the point: the answer lives in the workaround hours and the exit cost, so estimate those with real numbers before you choose. If you cannot estimate them, run the test plan below.

What custom costs you that a builder does not

Be fair to the builder side of the table. Custom code brings its own work. Someone has to host it, patch it, back it up and answer when it breaks at 7 in the morning. The client cannot edit it in a visual editor unless you build an admin screen for that, and that is extra cost. Hiring a developer later means they must read your code, so ask for plain documentation and a code repository (the place the code is stored with its history) in the client's name. Custom also takes longer to reach a first launch, so agree a small first phase with a date. If you cannot offer ongoing care yourself, name a partner who will, in the contract. Custom is an asset the client owns, but an unmaintained asset loses value fast.

A decision rule you can use today

  1. Count the signs from the list above that apply. Zero or one: use a builder.
  2. Two or three: run the two-week test (next section) before you decide.
  3. Four or more: quote custom, and say why in the proposal.
  4. Override: if the client has a hard exit or data-location requirement, that single sign decides it.

A two-week test plan with pass and fail rules

Do not argue about it. Prove it with the client's hardest feature.

  1. Pick the one feature that worries you most, for example the logged-in quote history.
  2. Build only that feature on the builder in at most 10 working hours.
  3. Write down what you used: the add-ons, the code snippets, the connectors.
  4. Pass if it works for a test user with real data, you can explain the setup to a colleague in 10 minutes, and the monthly extras stay under the figure you gave the client.
  5. Fail if it needs more than two embedded code snippets, depends on a connector you cannot trust to stay free or cheap, or you cannot get the data out in a file you can open.

A fail is cheap. It cost you ten hours and saved the client from a three-year problem.

How to say no without losing the client

Lead with their goal, not the tool. Say: "A builder will get you live in a week. It will also lock the booking data in their system and cost about X a month by year two. Here is the other route and what it costs." Show the three-year table with the client's real numbers, and the test result if you ran one. Offer a staged path: launch the marketing pages on a builder and build the one custom piece, such as the portal, as a separate app on a subdomain (an address such as app.clientsite.com). That keeps the first launch fast and puts the code where it needs to be. For pricing the retainer that follows, see how to price builds and retainers when AI cuts the hours, and for the wider buy-or-build question see build vs buy software and what custom software costs.

Who should skip custom

  • A business with a small budget and a standard brochure site.
  • A client who will not pay for care and updates. Custom code needs someone to look after it.
  • An agency with no developer and no partner. Do not promise what you cannot keep alive.
  • A team that cannot say what the custom piece must do, in one page. Write that page first.

Checklist before you quote

  1. List the features in three groups: pages, forms and content; accounts and data; connections to other tools.
  2. For each builder you consider, read its limits for the plan you would buy and its export page.
  3. Count the signs, and run the test if the count is two or three.
  4. Build the three-year table with your real rates and the vendor's current prices.
  5. Write the exit plan into the proposal and the contract.
  6. Decide who looks after the site and what that costs per month.

How I'd approach this

I would start every project on the assumption that a builder is fine, and make the project earn its way to custom. That keeps you honest, and it means the clients you do build custom for trust the reason. The sign I weigh most is exit: if the client must be able to take the work and walk, a platform with no export cannot be the answer, whatever else is true.

First week: take your three most recent proposals and count the seven signs on each. Then rebuild one of them as a three-year table with your real numbers. If the builder still wins, you have proof for the next client. If custom wins, you have a quote you can defend. When custom is the answer, Blur Studio builds the custom version: AI agents that use your business knowledge, connections to your CRM and calendar, finishing a half-built site or app, and ongoing care. See what that includes. A builder is still the right answer for many sites, and this page works either way.

For a closer look, see Ploy vs a custom Next.js site.

Questions people ask

Is a custom website always more expensive than a site builder?

At launch, usually yes. Hostinger's range puts custom builds at $5,000 to $30,000, while builders run from a few dollars to about $159 a month on Clutch's range. Over three years the gap can close or reverse when the builder needs add-ons, workarounds and a rebuild to leave. Run the three-year table with your own numbers.

What can a site builder not do well?

Heavy logic, user accounts with private data, deep connections to other systems, strict data location, very large sites and clean export. Each platform differs, so check the limits of the plan you would buy and the vendor's export page.

Can I export a website from Wix, Squarespace, Webflow or Framer?

As of November 2026, per their own pages: Wix says no, the site must run on its servers. Framer has no HTML export. Squarespace exports limited content as XML for WordPress. Webflow lets you export HTML, CSS and JavaScript on paid Workspace plans, but its pricing page says forms stop working and dynamic content (CMS items) must be exported collection by collection. See cancelling or exporting from a platform.

Can a hybrid work, a builder for pages and custom code for one feature?

Often yes, and it is a good default. Keep the marketing pages on the builder and build the logged-in or data-heavy part as a separate app on a subdomain, so each part uses the right tool.

How long does a custom build take?

It depends on scope, so ask for a fixed first phase and a written feature list. A small portal with a few screens takes weeks, not days, and a larger system takes months. Do not accept a date without a scope behind it.

What happens if the custom code is never maintained?

It slowly breaks as the tools and libraries it uses change, and security fixes are missed. Budget a monthly care fee, and make sure the client owns the code and accounts so another developer can take over. See who owns custom software code you paid for.

Does an AI tool change the answer?

It makes simple sites even cheaper to build, which strengthens the builder case for them. It also makes it faster to build a prototype of a custom piece, which helps the test plan. It does not remove the need for review and care. See finishing a half-built AI website.

How do I explain this to a client who just wants it cheap and fast?

Show the three-year table, name the one risk in plain words, and offer the staged path: launch on a builder now and add the custom piece when the need is proven. Put the exit plan in the contract either way.

More in this topic: Websites: agency delivery, quality control and margins (10 guides) →

Sources

Ready for the next step?

A 30 minute call. If you do not need me, I will tell you.

Ranjit Bhinge You talk to me, Ranjit. I reply within a few hours.

Not sure what you need? Tell me what is going on. I will point you the right way, even if it is not me.

Blur Studio · 30 minute call

Book a 30 minute call

You talk to me, Ranjit. If you do not need me, I will tell you.

  • We talk about your business and what is going on.
  • You leave with a clear next step, even if it is not me.
  • I reply to every message within a few hours.
Loading the calendar…