Free tool · No email required

WordPress, Webflow, Shopify or custom?

Answer what the thing actually has to do. The verdict appears as you answer, it names the single requirement that forced it, and it prints every rule underneath including the nine that raise nothing, so you can argue with the reasoning rather than with the answer. No email gate, no signup, no score. Nothing you type leaves your browser, because nothing is sent anywhere.

We are a custom development studio, so read the default state of this page first: with nothing ticked it returns a site builder, which is a thing we do not build and cannot invoice for. Three of the six possible answers here are ones we would hand to somebody else. A tool from a custom studio that always says custom is worth nothing, and you would be able to tell.

The tool needs JavaScript to run. The short version on paper: if nothing on your list needs a login, a domain model, live upstream data, writes into an internal system, real-time updates or an LLM feature, a site builder or WordPress will do it, and which of the two depends on whether a team publishes daily. If you sell a catalogue, use a hosted commerce platform unless your checkout has rules a hosted one will not run. Everything else on this page is the cost of ownership, which is the part worth reading with JavaScript on.

1. What is it, in one line

Used only to label the output. Nothing reads it and nothing is sent anywhere. Leave it blank if you like.

2. Who edits the content, and how often

Answer for the year after launch rather than for launch week. The first option is the one people are least willing to admit and it saves the most money.

NobodyA developer changes it, or it does not change.A couple of people, rarelyA few times a year, on the copy.A marketing person, weeklyPages and posts, one pair of hands.A team, daily, with reviewDrafts, scheduling, an approval step.

3. What the content actually is

The difference between the third and fourth options is whether the records have behaviour. If something has a status that changes and rules about what may happen next, it is a domain model and it is not content.

Pages and postsFlat. A page is a page.A handful of custom typesCase studies, services, people.Types that reference each otherOne edit fans out across many pages.A real domain modelStates, validation, derived values.

4. How much of it there is

Under 50 pagesA site, not a corpus.A few hundred itemsComfortable everywhere on this list.Thousands of itemsCheck them against published limits.Tens of thousands, or generatedOne page per row of a dataset.

5. Accounts and permissions

No logins for visitorsOnly the editors sign in.A members areaLog in, see what is otherwise hidden.Roles and per-record permissionsWho may see and edit which row.Separate customer organisationsIsolated data, their own users.

6. What is sold, if anything

Selling access to software is not this question. That is an account with a plan on it, and question 5 is where it gets decided.

Nothing is sold hereNo cart, no checkout.Enquiries, quotes, bookingsMoney changes hands elsewhere.A catalogue with a checkoutVariants, stock, tax, shipping.A recurring plan for those productsA box, a refill, a repeat order.A checkout with unusual rulesQuote to order, approvals, contract pricing.

7. Integrations with other systems

Tick a box only if the product is broken without it. Each one shows what it forces before you tick it, and two of the eight force nothing at all.

Forms into a CRM or mailing listA submission becomes a contact somewhere else.Forces: no floorAnalytics, tag manager, chat widgetThird-party scripts on the page.Forces: no floorA scheduled sync from an internal systemStock, prices or a calendar imported on a timer.Forces: a CMSData that must be current at request timeRead live on page load, not as of the last sync.Forces: a front end you ownSingle sign-on, SAML or OIDCPeople sign in with an account they already have.Forces: a CMSWrites into a system of recordThe site changes a row in the business's own software.Forces: a custom applicationUpdates without a page refreshA live queue, a collaborative view, a moving dashboard.Forces: a custom applicationAn LLM feature in the productReads your data, calls a model, costs money per call.Forces: a custom application

8. Compliance constraints

Tick these only where somebody has written the requirement down. Two of the six force nothing about your stack, which is worth knowing before a questionnaire is used to justify a rebuild.

Data must stay in a named regionA contract or a regulator says where the rows live.Forces: a CMSA retained trail of who changed whatSomebody will ask, and the answer has to be evidence.Forces: a CMSCard data touches your infrastructureNot a hosted checkout and not an embedded payment element.Forces: a custom applicationHealth data, or a signed agreement requiredA business associate agreement, or its local equivalent.Forces: a custom applicationA published accessibility standardWCAG 2.2 AA in a contract or a procurement pack.Forces: no floorEnterprise procurement reviewSOC 2, a pen test report, a subprocessor list.Forces: no floor

9. Expected traffic

This raises nothing, on purpose. It is here so the tool can say out loud that traffic almost never forces a platform, and so the running costs below can use your number.

Under 10,000 a monthAny answer here serves this.10,000 to 100,000Still not a platform question.100,000 to 1 millionA caching question, not an architecture one.Over 1 millionA CDN invoice, not a rebuild.

10. How bespoke the design must be

A good template, lightly themedColours, type, your logo.Your brand on a strong templateRecognisably yours, not from scratch.A bespoke designBuilt to a designer's file.Bespoke, with real motion and componentsInteraction, state, reuse with variants.

11. Who looks after it afterwards

Nobody yetYou will be buying maintenance.A marketing personContent, yes. Dependencies, no.A freelancer on callFine, until they are unavailable.One in-house developerOne codebase is affordable. Two is not.A development teamAnything on this page is maintainable.

12. Budget band

These are our published bands, not a field that sets your price. The tool uses them only to tell you when the verdict and the budget do not meet, and it says nothing at all when the verdict is one we do not sell.

Under $7,500Below our $7,500 minimum for a build.$7,500 to $15,000Inside our posted ranges.$15,000 to $28,000Inside our posted ranges.$28,000 to $45,000Inside our posted ranges.Over $45,000Above where most of our projects land.

Verdict for your project

Webflow, or another site builder

One subscription, no codebase, and a ceiling worth knowing the height of before you start.

We do not build this and we are not going to pretend otherwise. Nothing in your answers requires a custom development studio, so the useful thing this page can do is say so and stop.

What forced it

Nothing. Not one requirement you gave has a floor above a site builder, so there is no gate to clear and no case to argue. Buy the subscription, hire a designer, and put the difference into whatever part of the business is not this website. If somebody in the room still wants a build, that is a commercial argument rather than a technical one and it should be made as one.

Every rule, and whether it fired

No score and no weighting. Each rule sets a floor when it fires, a commerce rule changes the lane rather than the floor, and the verdict is the two combined. The last column is the running answer after each row, which is the entire arithmetic of this tool.

Every rule the tool evaluates, whether it fired for these answers, the floor it sets, and the running verdict after each rule.
Rule	Fired	Running
Nobody edits content after launchFloor: no floor	No	site builder
A team publishes daily, with reviewFloor: a CMS	No	site builder
Content types that reference each otherFloor: a CMS	No	site builder
A real domain model, not documentsFloor: a custom application	No	site builder
Thousands of content itemsFloor: no floor	No	site builder
Tens of thousands, or pages generated from dataFloor: a front end you own	No	site builder
Six figures a month or moreFloor: no floor	No	site builder
People log in to see somethingFloor: a CMS	No	site builder
Permissions that depend on the recordFloor: a custom application	No	site builder
Separate customer organisations with isolated dataFloor: a custom application	No	site builder
Enquiries, quotes and bookings onlyFloor: no floor	No	site builder
A real catalogue: variants, stock, tax and shippingLane: commerce platform	No	site builder
A recurring plan for those productsLane: commerce platform	No	site builder
A checkout with rules a hosted platform will not runLane: commerce platform, floor: a custom application	No	site builder
Forms into a CRM or a mailing listFloor: no floor	No	site builder
Analytics, tag manager, chat widgetFloor: no floor	No	site builder
A scheduled sync from an internal systemFloor: a CMS	No	site builder
Data that must be current at request timeFloor: a front end you own	No	site builder
Single sign-on against your identity providerFloor: a CMS	No	site builder
Writes into a system of recordFloor: a custom application	No	site builder
Something updates without a page refreshFloor: a custom application	No	site builder
An LLM feature inside the productFloor: a custom application	No	site builder
Data must stay in a named country or regionFloor: a CMS	No	site builder
A retained trail of who changed whatFloor: a CMS	No	site builder
Card data touches your own infrastructureFloor: a custom application	No	site builder
Health data, or a signed agreement requiredFloor: a custom application	No	site builder
A published accessibility standard you must meetFloor: no floor	No	site builder
Enterprise procurement reviewFloor: no floor	No	site builder
A bespoke design, built to a designer's fileFloor: no floor	No	site builder
Bespoke motion and custom interactive componentsFloor: a front end you own	No	site builder

What this verdict costs to run, and to leave

Four bills. One of them arrives monthly and the other three arrive when you are least able to absorb them. Third-party figures below are the vendors’ own published numbers, named where they are used and checked in September 2026. They move, and one of them moved this year.

What it costs to run, every month: Webflow publishes a Basic site plan at $15 a month billed yearly or $25 billed monthly, with a 300 static page limit, and a Premium plan at $25 a month yearly or $39 monthly that includes 20,000 CMS items and 40 CMS collections. Premium is the merger of the old CMS and Business plans, done in May 2026. Add a Workspace seat for whoever builds it. Figures as published by Webflow, checked September 2026: they restructured this once already this year.
What it costs to change your mind later: Webflow will export static HTML, CSS and JavaScript. It will not export the CMS-driven side, the hosting behaviour or anything behind a login, so treat the export as a design asset rather than as a working site. Rebuilding it as a marketing site in code is our $7,500 to $15,000 band over 2 to 4 weeks for up to 8 pages. Past 8 pages with a real content model behind it, it is a platform migration and it is priced like the WordPress to Next.js work we do most often, $16,000 to $28,000.
An estimate in US dollars, not a quotation; the quote you get after scoping is fixed in writing before kickoff.
Who you can hire for it: A very large pool of Webflow designers, which is the entire point of the platform and a real advantage. Almost none of them are engineers, so the failure mode is not that work gets harder when you hit the ceiling. It is that it stops, and the next quote you get is for a rebuild.
The lock-in that is not obvious on day one: The design is yours and the content model is not: your collections, references and interactions are expressed in Webflow's structures. The sharper one is that the site stops being served the month you stop paying. There is no self-hosted version and no read-only fallback, which makes this the only verdict on the page where a billing lapse is an outage.

Sources for the third-party figures on this page: Webflow plans and pricing and its May 2026 pricing update, Shopify’s pricing page, Advanced Custom Fields’ ACF PRO pricing, and Sanity’s pricing page. All read on 1 September 2026. None of them is an EdgeBrain price and none is a quotation from anybody. Read the vendor’s own page before you put a figure in a budget.

The five you did not get

Printed because the answer is only useful next to the alternatives, and because the recurring bill is the column nobody puts in a comparison post.

Every possible verdict, its recurring cost shape, and its least obvious form of lock-in.
Answer	What recurs	The lock-in
Webflow, or another site builderYours	Webflow: Basic $15 a month yearly or $25 monthly, Premium $25 yearly or $39 monthly with 20,000 CMS items and 40 collections. Published by Webflow, checked September 2026.	No self-host path. Stop paying and the site stops being served.
WordPress, with a theme and plugins	Core is free and GPL. The bill is licences: ACF PRO publishes $49 a year for one site, $149 for ten, $249 unlimited, renewing annually (Advanced Custom Fields, checked September 2026). Ten to fifteen such licences is ordinary.	The plugins, not the content. Each one is somebody else's release schedule and business model.
Shopify, with a theme	Shopify publishes Basic $25 monthly or $19 annually, Grow $65 or $49, Advanced $399 or $299, Plus from $2,300. Third-party payment fee 2 percent, 1 percent, 0.6 percent, 0.2 percent by plan. Checked September 2026. Apps are extra and add up.	The checkout is theirs, extended through their published points rather than edited.
Shopify behind a front end you own	The full Shopify plan and app bill, plus hosting and a build pipeline for the storefront. Headless adds a bill, it does not replace one.	A permanent second codebase, taken on to get around a theme. Most regretted choice here.
A headless CMS behind a front end you own	CMS subscription plus hosting. Sanity publishes a free tier at 20 seats and 10,000 documents, Growth at $15 per seat per month, plus per-unit overages on requests, assets and bandwidth. Checked September 2026.	Your own codebase. Untouched for two years it does not build, and no vendor is going to fix that.
A full custom application	Hosting, database, monitoring, email, CI, and auth if you buy it. No platform fee, and no platform absorbing anything either.	Your own decisions, with no vendor to blame and no vendor to fix it.

What we would charge, if anything

Not a line on our rate card, because we do not build these

If this is the verdict, hire a good Webflow designer. We would be the wrong studio and a more expensive one. If you would rather have the same site as code from the start, our marketing site band is $7,500 to $15,000 over 2 to 4 weeks for up to 8 pages, and $7,500 is our minimum for a build of any kind.

An estimate in US dollars, not a quotation; the quote you get after scoping is fixed in writing before kickoff.

This tool did not calculate that band and could not, because a platform is not a scope. For arithmetic with line items in it, use the MVP cost estimator.

When this answer changes

Not if. Vendor limits, vendor pricing and vendor feature sets are the inputs here, and all three are product decisions made by somebody else on their own schedule.

When your content count approaches the vendor's published limits. Webflow publishes 20,000 CMS items and 40 CMS collections on its Premium plan, and it restructured those plans in May 2026. Check them against your own numbers once a year, not once.
The first time somebody asks for a login. A gate is the cleanest exit signal a site builder gives you, and it arrives long before anything breaks.
Whenever the answer to any question above changes from a preference to a commitment made to a customer. That is the trigger. A quarter on a roadmap is not.

This picks the platform class, not the framework

“You need a custom application” is not the same sentence as “you need Next.js” or “you need Postgres”. This tool answers the first and refuses the second, because the second depends on who you can hire and what your data actually looks like. We have written that argument out separately in Postgres or Mongo for a new product and Vercel, AWS or a VPS.

What this cannot know

Whether a tick is real

A radio button is a label, not a specification. “Live data” means a five-minute cache to one person and a subscription over a websocket to another, and those are separated by a rung and a year of salary. The tool evaluates what you told it. A conversation evaluates the requirement, which is why the rule that forced your verdict is the first thing to pressure-test and the only one worth an argument.

Who is already in the building

A team with four WordPress developers and an agency retainer will build a better WordPress site than a better-chosen platform they have never used. That is not a technical argument and it is frequently the correct one. The tool sees your requirements. It does not see your org chart, your existing contracts or the person who will quietly refuse to learn the new thing.

What the vendors do next

Every limit and every price quoted here is a product decision somebody else makes on their own timetable. Webflow merged two plans and repriced in May 2026. Platforms add the feature that would have changed your answer, and they remove features people built companies on. Date the verdict, and put a reminder in the calendar rather than trusting a screenshot.

How to read the verdict

The verdict is a floor, not a recommendation. It names the least demanding platform class your stated requirements permit. It does not say that class is the one you should buy, and it never says a custom build is a good idea. If it returns WordPress, that means nothing on your list needs code from us. You are still free to commission a custom application. You will just be choosing to pay for one rather than being forced into one, which is a far better position to occupy when the quotes arrive.

Read the “true, and it changed nothing” list before you read the verdict. That list is where the argument inside your own building usually lives, because it names the things somebody was certain would settle it. Bespoke design is on it. Traffic is on it. An accessibility standard is on it. Then read the counterfactual line, which prices the single requirement that decided the answer, and go and ask whether anyone would miss it.

The four mistakes

Choosing the platform before the content model. Almost every regrettable stack decision on this list was made by picking a platform in a meeting and discovering the model in sprint three. The model is knowable on day one: write down the entities, the relationships between them, and which fields have rules. If a record has a status that changes and constraints on what may happen next, no CMS is going to hold it well, and finding that out early is worth more than any comparison table.

Buying flexibility you have no one to use.A custom build is not more flexible if nobody is employed to flex it. The platform you can actually staff beats the platform that is theoretically better, every time, and the honest version of the question is not “which is best” but “which of these can we keep alive for three years”. A codebase with no maintainer decays faster than a WordPress site with no maintainer, because at least the WordPress site has a vendor shipping patches.

Treating a compliance questionnaire as an architecture document. Two of the six compliance rules on this page raise nothing at all. A procurement pack asking for SOC 2 is a question about your operations, and an accessibility standard is a question about your markup and your testing. Neither is a reason to rebuild. The two that do bind, card data on your own infrastructure and a signature on a data agreement, bind hard and should be settled before anything else is ranked.

Underestimating the exit. Everybody prices the build and almost nobody prices leaving. The number that matters is not the subscription, it is what it will cost to be somewhere else in three years: the content export nobody has tested, the app data that lives in the app vendor rather than the platform, the redirect map for every URL you have ever ranked for. Ask for an export on day one, while you still have leverage, and see what actually comes out of it.

When the answer is WordPress and somebody senior wants custom

Do not argue with the preference. Ask what the custom build is for. If the answer is performance, the fix is caching and images and it costs a fraction of a rebuild. If it is “the site feels dated”, that is design and design does not force a rung on this page. If it is a plugin that keeps breaking, replace the plugin and count how many others are load-bearing while you are in there. If it is one of the rules that actually bound, the tool has already named it and you have the sentence you need.

The useful output of this exercise is rarely the verdict. It is the list of things that turned out not to matter, and the one thing that did.

Keep reading

Argue with the rule.

Send us the rule this tool says forced your answer and the reason somebody wrote it down. We will tell you whether it holds, including when the answer is that you do not need us.

Talk it through with us Price the build next