What it costs to build (or finish) an app in 2026

I'm not going to put a price on your app in this note. Nobody can do that honestly without knowing what the app has to do, and that missing piece explains most of the gap between the price ranges you'll find online. What I can do is show you the current numbers, each with a source and a date, and what pushes a project toward either end of them.

Start with the build. In a March 2026 GoodFirms survey of more than 100 software development companies (opens in a new tab), 66% said they typically charge $30,000 to $100,000 for a small to mid-sized project. That's the price of having an app built. Owning one costs more: Robert Glass's 2001 column in IEEE Software (opens in a new tab) put maintenance at 40 to 80 percent of a product's lifetime software cost, 60 percent on average.

If you already have half an app, from Lovable, Bolt, v0 or a freelancer who left mid-project, you're pricing a different job. Finishing gets its own section below.

How much does it cost to build an app in 2026?

Going by that GoodFirms survey, $30,000 to $100,000 is the common range for a small to mid-sized project. In the same release, companies building AI capabilities and advanced integrations often price projects at $50,000 to $125,000.

Two caveats. The firms doing the charging reported those numbers, and the survey doesn't define "small to mid-sized." I still start from it, because it's recent and it says where it came from.

Project complexity was the cost driver firms named most, at 95.3%, ahead of the technology stack at 79%. That matches what I see: the stack moves a price far less than what the app has to do.

Why are two quotes for the same app so far apart?

Because they're rarely quotes for the same app. One vendor prices a login, a few screens and a Stripe checkout button. Another prices user roles, subscriptions that handle failed renewals, an admin panel, a staging environment and store submission. Both call it an MVP.

Glass pointed at the deeper cause 25 years ago. Most estimates, he wrote, are made "before the requirements phase and thus before the problem is understood. Estimation therefore usually occurs at the wrong time." A price without a written scope is a guess, and two guesses can land a long way apart.

What do app developers charge per hour in 2026?

Hourly rates depend heavily on where the team sits. Accelerance's 2026 outsourcing rate guide (opens in a new tab), published in November 2025 from a survey of 60 development partners, puts them at:

• Latin America: $33 to $45 an hour for junior developers and $60 to $75 for senior, down 7.1% year over year
• Europe, including Central and Eastern Europe: $31 to $39 junior and $64 to $76 senior, down 4.4%
• Asia: $24 to $31 junior and $31 to $41 senior, down almost 8%

The guide also quotes Olivier Poulard, who runs global software engineering strategy at Accelerance: "Hourly rates are a poor measure of the true cost of software development."

For a US in-house comparison, the Bureau of Labor Statistics (opens in a new tab) puts the median annual wage for software developers at $135,980 as of May 2025. Over a 2,080-hour year, that's about $65 an hour in wages alone.

Wages aren't the whole cost of an employee. In the BLS release on employer costs for June 2026 (opens in a new tab), benefits made up 30% of what private-industry employers spent on compensation. If a developer's benefits follow that average, a developer on the median wage costs an employer roughly $194,000 a year, or about $93 an hour, before recruiting, equipment and management time. It's a rough figure: the 30% covers every private-industry job, not just software.

I haven't found a good public source for US agency rates. The ranges I've seen come from agencies selling at them, so get rates in writing and compare.

Freelancer, agency or in-house?

The rate also hides what's included. A freelancer's rate usually buys one person writing code; an agency's tends to fold in project management, design and QA. In-house has no markup, and it's the hardest of the three to unwind if the hire doesn't work out.

Whichever you pick, ask who owns the work nobody put on the invoice: testing, deployment, monitoring, and answering the customer who says they were charged twice.

What drives the cost of building an app?

Before anyone can price your app, they need answers to four questions.

Who uses it, and what can each role do?

An app where one kind of user sees their own data is a different job from one where customers, staff and admins each see different things and money moves between them. Every role adds screens, permission rules and test cases.

The easiest piece to forget is the admin side. Someone will need to issue a refund, fix a customer record or block a spammer, and those screens are real work. Regulated data (health records under HIPAA, EU personal data under GDPR) adds access controls, audit logs, vendor agreements and data-retention rules on top.

Web, iOS or Android?

Web only is the cheapest start, and for many B2B products it's enough. Native mobile adds store accounts, review cycles, device testing and a release build for each store.

React Native with Expo, or Flutter, lets one codebase ship to both stores. We built SkillSchedule's iOS and Android app on Expo, alongside its web portals. That saves writing the app twice, though each store still needs its own build and review.

Is it a checkout or a billing system?

"Add Stripe" is one line in a feature list and one of the widest ranges in any estimate. A one-off checkout is small. Subscriptions aren't: Stripe's docs on subscription webhooks (opens in a new tab) say the integration "requires a destination that handles webhook events, because most subscription activity happens asynchronously."

Renewals, failed cards, refunds and disputes arrive as events your app has to receive, verify and act on. Skip the risk events and, in Stripe's words, your app "won't learn about disputes or fraud warnings until you check the Dashboard."

A checkout that only confirms the payment when the customer lands back on your page demos perfectly. It never hears about the refund, the dispute or the renewal that fails a month later.

What outside services does it call?

Every outside service costs twice: the work to connect it properly, then a bill that grows with use. Twilio SMS, Google Maps Platform and AI model APIs are priced per message, request or token. At demo volume that rounds to zero. At a few thousand active users it's a line on your P&L.

What hidden costs should you budget for?

These are the ones that tend to arrive after the build quote is signed.

Free tiers that aren't for businesses

Before launch, read the terms of every free tier the app runs on. Check whether commercial use is allowed and what happens to a project that sits idle, and budget for the paid plan if either answer is a problem.

Security left at its defaults

The first thing I check is who can read which rows in the database. Row-level access rules touch every table, which is why adding them after launch means reopening every query. I've written up that gap and the other common ones, with a self-check you can run in ten minutes, in security gaps in AI-built apps.

No staging, no tested rollback

An app that lives only inside the tool that built it usually has a single environment, so every test runs against the database your customers use. Reverting the code in the builder doesn't undo what a bad release did to your data. Set up staging and a rollback you've tested before launch, while nobody is waiting on you.

Store review and account deletion

If you're shipping to the stores, leave room in the launch date for a rejection. Apple said in May 2026 (opens in a new tab) that App Review evaluated more than 9.1 million submissions in 2025 and rejected over 2 million, including nearly 800,000 app updates and over 443,000 submissions for privacy violations. Each rejection means a fix and another round of review.

One requirement I'd put in the scope from the start is account deletion. Since June 30, 2022 (opens in a new tab), Apple has required apps that support account creation to let users start deleting their account inside the app. Google Play's policy (opens in a new tab) asks for an in-app path plus a web page where users can request deletion of their account and data. Neither is hard to build, and both are easy to leave out of a quote.

Is it cheaper to finish a half-built app or start over?

It can go either way, and anyone who answers before looking at the code is guessing.

When finishing costs less

A working prototype holds decisions you already paid for: which screens matter and what customers reacted to. If the data model is sensible and what's missing can be added without moving what's there, production work can go underneath the screens you have.

The case against starting over is old and still holds. In a 2000 essay (opens in a new tab), Joel Spolsky called rewriting code from scratch "the single worst strategic mistake that any software company can make," because old code carries fixes for problems that only real use turned up. "When you throw away code and start from scratch, you are throwing away all that knowledge." A prototype holds less of that knowledge than a mature product does, but what it holds, you paid for.

When finishing costs more

Glass wrote that the one task maintenance adds to development, "understanding the existing product," is the dominant maintenance activity, at roughly 30 percent of maintenance time. Code nobody on the team wrote, including code a model wrote, carries that cost on every change.

In Stack Overflow's 2025 Developer Survey (opens in a new tab), the most common frustration with AI tools was "AI solutions that are almost right, but not quite," named by 66% of those who answered, and 45.2% said "debugging AI-generated code is more time-consuming." Plenty of half-built apps are in that state: they work until a real user takes a path nobody tested.

Finishing also costs more when the foundation has to change under everything: access rules missing across the schema, several customers' data mixed in the same tables, or payments built on a builder's own payment feature, where moving to your own Stripe account can mean a rebuild.

If the app will hold other people's money (payouts, escrow, a marketplace, beyond simply taking payments), keep several companies' data or run a regulated workflow, plan to keep the screens as the spec and rebuild the foundation. The full list of signs is in where AI app builders stop.

How do you size the cost of finishing an app?

Finishing is priced by the gap list, not by the screens. The usual items:

• Stripe webhooks for renewals, refunds, disputes and failed cards
• Access rules on every table, enforced by the server
• Separate staging and production, with a rollback you've tried
• Secret keys moved out of the browser
• Error alerts that reach a person, and a backup you've restored
• Store builds, review fixes and account deletion, if you're going mobile
• The domain, repo, database and Stripe account in your company's name

For the complete list, work through the Production Gap Checklist. Put hours against each gap, add them up and multiply by a rate.

An illustration, not a quote: two senior developers for 12 weeks is 960 hours. At Accelerance's Latin America senior rate of $60 to $75 an hour, that's $57,600 to $72,000 (my arithmetic). The 12 weeks is an assumption to show the math; your gap list sets the real number.

How much does app maintenance cost per year?

The rule of thumb I plan with is 15 to 20 percent of the build cost a year. Treat that as a floor for a quiet year, then add store fees, forced platform updates and the hours someone spends patching dependencies and watching the error logs.

Where does 15 to 20 percent come from?

It's a planning number, not a measurement. I've looked for a study behind it and haven't found one.

Glass's figure measures something different: maintenance as a share of everything spent over a product's life, 40 to 80 percent, 60 on average. He also split it. Roughly 60 percent of maintenance is enhancement and about 17 percent is error correction. Maintenance, he wrote, "is largely about adding new capability to old software, not about fixing it."

The two numbers fit together. At 20 percent of the build cost a year, maintenance makes up 60 percent of everything spent by about year seven and a half; at 15 percent, by year ten.

Forced updates and store fees in 2026

Even an app nobody touches needs work before it can ship its next update. On iOS, Apple's February 2026 notice (opens in a new tab) says that from April 28, 2026, apps uploaded to App Store Connect must be built with the iOS 26 SDK or later.

On Android, Google Play's target API level rule (opens in a new tab) says that from August 31, 2026, new apps and updates must target Android 16 (API level 36), and existing apps targeting below Android 15 stop being offered to new users on newer devices. Meeting either rule means dependency upgrades and a round of retesting, planned or not.

Store fees changed this year too. Google Play's new fee structure (opens in a new tab) began June 30, 2026 in the US, UK and EEA. It charges a 10% service fee on your first $1M in annual earnings and on auto-renewing subscriptions, plus a 5% billing fee if you use Google Play's billing. That's 15% in total, unless you use alternative billing or a web link, which the new structure exempts from the billing fee.

On iOS, Apple's Small Business Program (opens in a new tab) charges a 15% commission on paid apps and in-app purchases if you enroll and earned up to $1M in proceeds the year before; new developers qualify too, and the standard rate applies otherwise. If you sell subscriptions in the app, put those percentages in your pricing model now.

Common cost questions

How much does an MVP cost in 2026?

In GoodFirms' March 2026 survey, 66% of development firms said they typically charge $30,000 to $100,000 for a small to mid-sized project, a reasonable starting range for an MVP. Payments, several user roles and an admin panel each push it up.

How much does it cost to finish a Lovable or Bolt app?

It depends on the gap list. The 12-week illustration above comes to $57,600 to $72,000; an app that holds other people's money, several companies' data or a regulated workflow costs more, because its foundation gets rebuilt.

Do AI coding tools make app development cheaper?

For parts of the work. In GoodFirms' 2026 survey, 61% of firms expected AI to cut project budgets by 10 to 25%. What I don't see getting cheaper is checking the code before real users touch it.

How do you get a real number for your app?

Give every vendor the same short brief and the quotes become comparable. Answer the four cost-driver questions above, say what data exists already and whether any of it is regulated, and link to whatever is built so far.

Then ask how it's priced. A fixed price puts the risk of a wrong estimate on the vendor; time and materials leaves it with you. Either can be fair. A fixed price with no written scope is the one to walk away from.

We don't publish a price list at Synaptech, for the reason I gave at the top: without a scope, any number I printed here would be wrong for most readers. What we do instead is write the scope with you, free, in the Production Gap Review.

You send what you have: a Lovable or Bolt share link, a repo, a project someone left half-done, or a few lines about the idea. We talk it through for 30 minutes, and then I write down every gap between that and a product real customers can use. It's the brief I described above, and you can hand it to any vendor you're weighing, us included.

If we're the team for the job, our quote is a fixed price against that written scope. Finishing a prototype or building a first version is our Launch Engineering work; a platform that needs its foundations built properly from day one is Product Engineering.

Whoever you hire, compare prices only once every vendor is quoting the same written scope. Before that, the cheapest quote is often just the one that left the most out.