Skip to content
Rivl
15 November 2026Cost10 min

How much does an MVP cost, and what really moves the number

There is a floor you can look up in ten minutes and a build cost nobody can quote without a scope. Confusing the two is how an MVP ends up costing twice: once to make, and once to throw away.

The question arrives as a request for a figure, and the honest answer is that the figure is the output of three decisions you have not made yet. That sounds like an evasion. It is not, and the proof is that the same feature list, built for two different answers to those three questions, differs in cost by a factor of about four.

What can be priced precisely, right now, is the floor: the fees that exist whatever you build. Almost nobody includes them, and on a small MVP they are not a rounding error.

The sections of this note in order: the floor you can look up, the three variables that actually move the build, the four shapes of MVP compared, why the cheapest version is often the most expensive, the costs that arrive after launch, what to ask for instead of a number, and honest limits.
The third and fourth sections are the ones that change what you commission.

The floor you can look up today

If the MVP is a mobile app, distribution is not free and the terms are published. Apple's Developer Program is a 99 dollar annual membership, and it renews. The free registered account lets you build and test on your own device, but it does not include App Store Connect, TestFlight or signing certificates, which means it cannot ship anything to anybody.

Google Play is cheaper and stranger. The Play Console charges a 25 dollar one-time registration fee, with no renewal. The cost that catches people is not the money. Google requires identity verification during setup, and it states that the registration fee is not refunded if the information is judged invalid. Personal accounts carry two further conditions: they must complete specific app testing requirements before the app can be made available, and the owner must verify access to an Android device through the Play Console mobile app before publishing.

Read that as a schedule risk rather than a cost. A team that plans a four week MVP and discovers the testing requirement in week four has not lost 25 dollars, it has lost its launch date. Register the accounts in week one, before any code exists, because the verification steps run on somebody else's clock.

Web-only MVPs avoid all of that, which is a real argument for starting on the web even when the product is eventually a phone app. The running costs are a separate question and they are small at the start, covered in what a web app costs per month to run.

The three variables that actually move the build

Everything else about the price comes down to three questions, and none of them is about the day rate.

  • How many kinds of user. One role is a product. Two roles that see different things, with permissions between them, is closer to two products, because every screen now has a second version and a rule about who gets which.
  • Whether the data starts somewhere else. A system that begins empty is cheap. A system that has to import a decade of spreadsheets, or stay in step with an accounting package, carries work that has nothing to do with the feature list and usually exceeds it.
  • Whether it has to be correct. A tool that is merely useful can be wrong occasionally and corrected by a human. Anything touching money, stock, payroll or a regulator cannot, and the difference is in the testing, the audit trail and the error handling rather than in the screens.

That third one is the most commonly mispriced. Two products can look identical in a demo and differ by months, because one of them is allowed to fail quietly and the other is not. If you are comparing quotes and one is much lower, this is usually where the difference went.

Four shapes of MVP, and what each one buys

We are not going to print a price range, for a reason given in the limits below. What is useful instead is the shape, because the shape decides the order of magnitude and you can pick it yourself.

ShapeWhat it isWhat it provesRough effort
Clickable prototypeReal screens, no working dataThat people understand it and want itDays
Single path web appOne role, one job, done properlyThat the job gets done and people returnTwo to six weeks
Internal toolReplaces a spreadsheet, with roles and historyThat the process survives being made explicitFour to ten weeks
Mobile or two-sidedTwo roles, store review, notificationsAlmost nothing a web version could not prove soonerMonths

The first row is the one most often skipped and the one with the best return. A prototype that somebody clicks through finds disagreement that a requirements document hides, and it finds it before anything is expensive to change. The argument in full is in a clickable prototype before development.

The second row is where most honest MVPs live. What fits in that window, and what quietly eats it, is set out in an MVP in one week, which is the compressed version of the same arithmetic.

Why the cheapest MVP is often the most expensive

There are two ways to build something cheaply. You can reduce what it does, or you can reduce how well it is built. The first is a strategy. The second is a loan at an interest rate nobody quotes you.

The failure mode is specific and common. The cheap build is commissioned to test an idea, the idea works, and now the thing that proved it is also the thing in production. It has no tests, one developer understands it, the data model assumed one role, and the next feature costs more than the original build did. At that point you are not extending an MVP, you are paying for it a second time while running it.

The way out is to decide, before the work starts, which of those two things the MVP is. A throwaway MVP should be genuinely throwaway: shortest path, no migration plan, explicitly scheduled for deletion. A foundation MVP costs more per feature and fewer features fit, and that is the trade. What does not work is commissioning the first and planning to treat it as the second, which is the default and the reason the rebuild versus refactor question arrives so much earlier than anyone expects.

The costs that arrive after launch

An MVP budget that stops at launch is not a budget, because the first three months after launch are when the real requirements show up. Expect three lines nobody puts in the original number: the fixes that come from actual use, the hosting and third party services once there is traffic, and the small changes that are the whole point of having shipped.

Those are predictable in shape even when they are not predictable in size, and the ranges plus what drives them are in software maintenance cost after launch. The short version is that treating year one as build plus nothing is the single most reliable way to be surprised.

What to ask for instead of a number

If you want a figure you can rely on, the fastest route is not to ask more firms for quotes. It is to answer the three questions above in writing, in a page, and send the same page to everyone. Quotes that differ by a factor of three usually differ because each firm guessed differently at what you meant, not because one of them is cheap.

What that page needs to contain, and the two things that cause most of the discovered work, are in how to scope a software project before anyone writes code.

If the MVP is a shop rather than software, the cost question has a much more concrete answer, because the platform fees, payment rates and delivery costs are all published. Our Arabic note on the itemised cost of a complete online store in Egypt works through those line by line. It is written in Arabic and is about the commercial setup rather than a custom build, which is the half this note leaves out.

Honest limits

The four shapes of MVP from the table in this note, in order of effort: a clickable prototype taking days, a single path web app taking two to six weeks, an internal tool taking four to ten weeks, and a mobile or two-sided product taking months.
Effort, not price. The shape decides the order of magnitude.

We have not given you a currency figure, and that is deliberate. We would have to invent a day rate and a market, and any number we printed would be quoted back at us by someone whose project is nothing like the one we imagined. The store fees above are published and checkable. A build price is not, until there is a scope.

The effort column is our own experience, not a survey. It comes from projects we have run, which is a small and biased sample. It is useful for choosing a shape and useless as a benchmark, and we have no published dataset for MVP cost in Egypt or the Gulf that we would be willing to cite.

Store fees and requirements change. The Apple and Google figures above were read from their own pages on 15 November 2026. Check them rather than trusting this note, particularly the Play Console testing requirement, which applies to personal accounts and has changed more than once.

And sometimes the answer is not to build anything. If an existing product does eighty per cent of the job, the MVP that matters is a two week trial of that product with real data in it. Custom software versus off the shelf is the decision to make before this one, and it is the cheaper question.

Describe it. We build it.

Seven or twelve days, pay on delivery, a year of maintenance included. Bring the problem, not a spec.

Book a meeting

Read next