App Development Cost in 2026: $8K–$500K Guide

Published: 5 August 2026 · Last reviewed: 5 August 2026

The price table, first

A note on where these numbers come from, before you use them. The tier prices, hour estimates and feature costs throughout this guide are practitioner estimates — our own observed market pricing, not audited survey data. Where a claim rests on published research or a vendor's own published price, it is linked inline. Where it is our judgement, this guide says so. Treat the two differently.

Everything below explains how to work out which row is yours, what each line item really costs, and why the same brief will come back quoted at $22,000 and $190,000 from two firms who are both telling the truth.

Why the same app gets quoted four times apart

This is the question underneath "how much does it cost to build an app," and most guides skip it. There are five real reasons, and only one of them is anyone being dishonest.

1. Hourly rate varies by a factor of five or more, legally. Compiled 2026 rate surveys from Second Talent (updated May 2026) and Uvik (April 2026) put mid-level mobile developers at roughly $17–$50/hour in South and Southeast Asia, $30–$68 in Eastern Europe, and $72–$100 in the United States, with senior rates running above the top of each band — Uvik lists senior India at $50+ and senior Poland at $65+.

Two cautions on those figures. First, they are developer rates, not agency bill rates: a US studio quoting $150–$200/hour is charging for a team and a business, not one person's time. Second, they are self-reported by the platforms compiling them.

For context on the floor under US rates: the U.S. Bureau of Labor Statistics puts the median annual wage for software developers at $133,080 as of May 2024, the most recent figure in the Occupational Outlook Handbook. That is about $64/hour in raw salary before benefits, taxes, management, sales cost, bench time or profit. A US shop billing $150/hour is not gouging; it is running a normal services business at a normal margin.

2. "The app" means different things to different estimators. One firm heard "marketplace" and scoped Stripe Connect, KYC, dispute handling, payouts and a tax form pipeline. The other heard it and scoped a listings screen with a contact button. Both quoted honestly. Nobody wrote the scope down.

3. Fixed-price quotes carry a risk premium — and they should. A fixed bid means the vendor absorbs the overrun. Historically, that overrun is not small. Flyvbjerg and Budzier's study of 1,471 IT projects, published in Harvard Business Review in September 2011, found an average cost overrun of 27% — but one in six projects was what they called a Black Swan, overrunning by an average of 200% on cost and nearly 70% on schedule. That distribution has held up; the same research group revisited IT cost risk across 23 project types in "The Uniqueness of IT Cost Risk," published online in Project Management Journal in July 2025 (2026 issue), concluding that IT cost risk has a fatter tail than any other project type they measured and cannot be predicted by conventional methods. A vendor pricing a fixed bid is pricing that tail. In our experience the premium runs 20–40% over the same work on time-and-materials — that figure is our observation, not a surveyed number. It is the cost of moving risk off your balance sheet, and it is often worth paying.

4. Some quotes exclude things you will need anyway. Design, QA, project management, App Store submission, backend infrastructure, and the first three months of bug fixes are all commonly quoted separately or not at all. A $22,000 quote that excludes all six is a $40,000 project.

5. Estimation is genuinely hard, including for the estimator. The most-cited data point here needs updating, because the picture changed in 2026. METR's randomized controlled trial, published July 2025 (arXiv:2507.09089), had 16 experienced open-source developers do 246 real tasks in mature repos, half with AI coding tools allowed. They expected to be 24% faster with the tools. Afterwards they believed they had been 20% faster. Measured, they were 19% slower.

METR has since superseded that result. In February 2026 it published data from a follow-up experiment run with late-2025 tooling and estimated an 18% speedup among returning developers and a 4% speedup among newly recruited ones — while stating plainly that this new data is "an unreliable signal," because developers who most expected to benefit from AI increasingly declined to participate at all, 30–50% of participants avoided submitting tasks they thought AI would help with, and participant pay had dropped from $150/hour to $50/hour. METR is redesigning the experiment. It now labels the original 19%-slowdown finding as historical.

So: do not quote the 19% slowdown as current. What survives is the perception gap the original study documented — developers were wrong about their own speed in both directions, by a wide margin. When a vendor tells you modern tooling has cut their estimate in half, ask what they measured. On greenfield code with no legacy context, the speedups are real. On a mature codebase with real constraints, the best available evidence is still noisy, and it compresses much less than the pitch suggests.

Self-diagnosis: which tier are you actually in?

Answer these honestly. Score one point per "yes" in the Tier 2 column, three per "yes" in the Tier 3 column.

Score 0–1: Tier 1. Build the cheapest credible version. Anything you spend above $35,000 right now is buying certainty you have not earned yet.

Score 2–5: Tier 2. This is where most funded startups and mid-market internal tools actually live, and where the $40K–$150K range is honest.

Score 6+: Tier 3. Stop reading cost guides and get an architecture discovery engagement first — in our market that runs roughly $15,000–$40,000 for 3–6 weeks. It will feel like paying for slides. It is the cheapest insurance available against the fat tail Flyvbjerg's data describes.

The most expensive mistake in this whole document is a Tier 3 problem bought at a Tier 1 price. The build finishes, technically works, and then cannot be extended — so you pay again. The second most expensive is a Tier 1 problem bought at Tier 2 prices, which is just an expensive way to learn nobody wanted it.

MVP vs full product: the split that saves the most money

An MVP is not "the same app with fewer features." It is a different artifact with a different job: answer one question cheaply.

Two independent survey sources land in roughly this territory. GoodFirms' developer survey puts a simple app at three to seven weeks and a medium-complexity app at $61,000–$69,000. Clutch's directory data — updated 5 August 2026 — shows reviewed mobile app projects typically costing $10,000–$49,999, with a reported average of $90,780 and a typical timeline near 11 months. Note what the gap between "typical" and "average" tells you: a small number of very large builds drags the mean far above the median. The loud end of the market is not the typical end.

The honest low end: a competent solo freelancer, working from a clear spec, using a managed backend and a cross-platform framework, can deliver a genuine working v1 for $8,000–$25,000. This undercuts nearly every agency, including ours. It is a real option and you should consider it. The trade-offs are concrete: no redundancy if that person disappears, no design or QA specialist, no project management, and code quality you cannot assess until it is too late. Arc's freelance rate data puts average and median freelance mobile developer rates at $61–$80/hour — note that Arc does not date this figure, so treat it as indicative — which means $20,000 buys roughly 250 to 330 hours, or six to eight focused weeks. That is enough for a real Tier 1 app and nothing more.

Web vs native vs cross-platform

The cost multipliers above are our own estimates, not survey figures.

Practical guidance: if you cannot name a specific capability that cross-platform cannot do for you, you do not need native. And if your app is fundamentally a set of forms and lists behind a login, seriously price the web version before you commit to stores. Web strips out App Store review cycles, the $99/year Apple Developer Program fee, and — for anything transactional — the store commission entirely.

Feature-by-feature: what individual things cost

These are our own build estimates at a blended $85/hour, roughly a good Eastern European or hybrid US/offshore team rate. They are not survey data. Multiply by 1.7–2.2 for a US-onshore team; by 0.5–0.7 for South Asia.

One 2026-specific note on that last row. The European Accessibility Act became enforceable on 28 June 2025. Consumer-facing apps offered in the EU must meet EN 301 549 v3.2.1, which incorporates WCAG 2.1 Level AA. Penalties are set nationally and vary widely: per Level Access's compliance overview, maximums run from around €40,000 per violation in Italy through €100,000 in Germany to roughly €600,000 in Spain and €900,000 in Sweden. Microenterprises — under 10 employees and under €2m turnover or balance sheet total — are exempt from the services obligations, but not from the product obligations, and the exemption lapses the moment you cross the threshold. Watch the standard, too: EN 301 549 v4.1.1, expected during 2026, moves the bar to WCAG 2.2.

Published estimates for the cost of accessibility are all vendor-produced and vary by an order of magnitude, so treat any precise percentage with suspicion. The consistent, defensible finding across them is directional and worth acting on: building accessibly from the start adds a modest premium, and retrofitting costs several times more. If you have any EU consumer exposure, put it in the original scope.

Integrations: where estimates go wrong

Integrations are priced by whose API it is, not by what it does. The ranges below are our own estimates.

The rule: an integration's cost is set by the slowest human on the other side. Before you accept any integration estimate, ask the vendor whether they have sandbox credentials in hand. If the answer is no, that line item is a guess.

Backend, infrastructure and running cost

Development cost is one-time. Infrastructure is forever, and it is the line item founders forget when they model unit economics. Build costs and monthly ranges below are our estimates; the vendor list prices are linked.

On BaaS specifically: do not assume the managed-Postgres option is automatically cheaper. One January 2026 cost comparison found Firebase cheaper than Supabase at small scale — around $7/month versus Supabase's $25 base at 5,000 users — with the crossover where Supabase becomes cheaper landing somewhere around 30,000–50,000 monthly active users. Model your own read patterns before you pick.

Add these recurring costs regardless of architecture, because they are near-universal and rarely quoted. Vendor plan prices change often; check before you budget.

Store commissions: read this section twice, it changed

Apple's $99/year and Google's one-time $25 are the trivial part. The store commission is not — and the 2026 position is genuinely unsettled, so any guide quoting a flat 30% is out of date.

The baseline rates still exist: Apple and Google each take 30% of digital-goods revenue, dropping to 15% for smaller developers. The structures differ. Google applies the 15% rate to the first $1M of earnings each calendar year and 30% above it; Apple's Small Business Program is a threshold you qualify for based on the prior year's proceeds and fall off entirely once you cross $1M, as RevenueCat explains in detail. Apple also charges 15% rather than 30% on subscriptions after a subscriber's first year.

The US position has diverged sharply. Following the Epic litigation, US developers can now steer users to external payment options on both stores:

• Apple. The Ninth Circuit affirmed the civil contempt finding against Apple in December 2025 and largely upheld the resulting remedies, though it vacated the blanket commission ban and remanded. The Supreme Court granted certiorari on 30 June 2026. In the meantime, Apple cannot collect a commission on qualifying US external-link purchases. A ruling is not expected before 2027.
• Google. Under the October 2024 injunction, Google's US external content links and alternative billing programs went live on 9 December 2025. Epic and Google jointly withdrew a proposed alternative settlement on 15 July 2026, leaving the original injunction governing. Rival app stores began appearing inside Google Play from 22 July 2026, and enrolled developers must report external transactions and pay the applicable service fees from 1 October 2026.

If you sell digital subscriptions, model the in-app rate and the external-link rate for every market you sell in, and check the current position before you set your price — this is the fastest-moving number in the whole guide.

One more 2026 practicality: if you register a new personal Google Play developer account, Google requires a closed test with at least 12 testers opted in continuously for 14 days before you can go to production. The requirement applies to personal accounts created after 13 November 2023. Organization accounts verified with a D-U-N-S number are exempt — but that verification itself commonly takes two to four weeks, so it is not automatically the faster route. Either way, that is calendar time, not money, and it has blown more launch dates than any technical problem.

Design and QA: the two line items people cut, and shouldn't

Design. In our experience a real design phase is 12–20% of build cost. For a Tier 2 app: 120–250 hours at $70–$150/hour, so $10,000–$35,000. That covers user flows, a component library, both platform variants, empty/error/loading states, and a handoff file engineers can actually build from. Skipping the design system is what produces an app where every screen looks like a different app, and where every new feature costs more than it should because there is nothing to reuse.

QA. Testing typically runs 20–30% of total development cost in the projects we price — again, our figure, not a surveyed one. For a Tier 2 app that means $10,000–$40,000, split between manual device testing (there are still a dozen screen sizes that matter) and an automated regression suite. If you want an external benchmark for the direction rather than the number, Capgemini and Sogeti's annual World Quality Report is the standard reference for enterprise quality-engineering spend, though its detailed budget figures sit behind registration.

The cheapest legitimate way to cut QA cost is to reduce surface area, not to reduce testing. Fewer platforms, fewer user roles, fewer configurable options. Cutting the QA line while keeping the scope just relocates the cost to your post-launch budget, at a significantly worse rate.

Post-launch maintenance: the number almost every guide understates

Here is the sentence most cost guides bury: your app costs money every year, forever, and the first year is not the cheapest.

The industry benchmark most often cited is 15–20% of the original build cost, annually, attributed to CISQ. See Cleveroad's breakdown, which cites CISQ directly, or Savi's summary, which reaches the same 15–20% figure from general industry benchmarks and adds a more striking one: IEEE research putting roughly 60% of total software lifetime cost in the maintenance phase and only 40% in initial development. That 15–20% is a reasonable floor. It is not the number to plan with.

Here is what the money actually goes to. The percentage splits below are our own allocation, offered to show where the money goes rather than as measured data:

Plan on 20% minimum for a maintained-but-static app, and 30–40% for one that is still improving. A $100,000 build is a $20,000–$40,000-per-year commitment. Over a five-year life, maintenance typically exceeds the original build cost — which is exactly what the IEEE 60/40 split above describes. Poorly built or inherited legacy code pushes the annual figure higher still, before any new features.

If you cannot fund year two, you cannot afford year one. This is the single most useful sentence in this guide.

Offshore vs onshore, honestly

Headline rates below are drawn from the Second Talent and Uvik 2026 surveys and reflect mid-level to senior developers. Both are self-reported. The "effective cost" column is our own adjustment.

\* Effective cost adds the overhead offshore buyers underestimate: your own time managing, rework from misread requirements, and the async delay tax. Budget 15–30% on top of the headline rate. It is real and it is not the vendor's fault.

What actually goes wrong offshore, in order of frequency:

1. Requirements ambiguity resolved without you. A US team asks a clarifying question in the hallway. An offshore team makes a reasonable assumption and you see it in the demo eight days later. The fix is written specs and daily written check-ins — not more calls.
2. The senior you interviewed is not the person coding. Ask, in writing, for named individuals with CVs, and a contractual right to be notified of substitutions.
3. Timezone compounding. Every round trip costs a day. Five clarifications on one feature is a lost week. Overlap hours matter more than rate.
4. No IP assignment clause that survives the jurisdiction. Have a lawyer read the contract. Not the vendor's lawyer.

When offshore is clearly right: well-specified work, Tier 1 and much of Tier 2, long-running teams you will keep for 12+ months, and any case where the rate difference funds a longer runway.

When onshore is clearly right: regulated data, work requiring constant business-context judgment, short high-stakes timelines, and any project where you personally cannot commit 8+ hours a week to written communication.

The blended model — an onshore lead or product owner plus an offshore delivery team — captures most of the savings and most of the accountability. It typically lands at a $70–$110 blended rate and is what a lot of mid-market agency pricing quietly is already.

What the market actually charges: vendor types side by side

The bands below are our own observations of the market, not surveyed data. For an external anchor, Clutch's directory reports mobile app projects typically costing $10,000–$49,999 with an average of $90,780.

Disclosure: this guide is published by ReactivVibeAI Studio. For comparison against the bands above, our own published Tier 2 pricing is $16,500 for a single developer and $29,500 for a team — at the low end of the small-studio band, because we scope tightly and decline Tier 3 work. If your build needs HIPAA, offline sync or a legacy ERP integration, we are the wrong vendor, and a mid-size agency's $200K quote is the honest one.

On the no-code row: the platform fees are small but permanent. FlutterFlow's plans run $39/mo Basic, $80/mo Growth and $150/mo Business as of March 2026, with Growth and Business priced per seat and backend hosting (Firebase or Supabase) billed separately; Bubble's Starter plan is around $29/mo billed annually or $32/mo month-to-month for web only, with hosting included. For a Tier 1 internal tool or a marketplace test, no-code is frequently the correct answer and everyone in this industry knows it. The honest warning is about the exit: migrating off a no-code platform later is expensive, because you keep the requirements and throw away the implementation — budget for a substantial fraction of a fresh build rather than a cheap port. Choose no-code when you expect to validate and rebuild, not when you expect to validate and grow into it.

Three worked budgets

These are illustrative reconstructions of typical projects, not quotes.

A. Local services booking app (Tier 1) — $27,400

B. B2B field-operations app with web admin (Tier 2) — $126,000

C. Regulated health marketplace (Tier 3) — $453,000

Notice what scales fastest between B and C. It is not the app screens. It is architecture, compliance, QA and integrations — the invisible half.

Hidden costs nobody quotes you

Our own estimates, from projects we have watched land.

How to read a quote

Five questions that separate a real estimate from a hopeful one:

1. "What's explicitly out of scope?" A vendor who cannot list five excluded things has not thought about it.
2. "Show me the assumption list." Every estimate rests on assumptions. If they are not written down, they are not shared.
3. "Who specifically is on this project, and what's your bench policy?" Names and CVs, plus what happens if someone leaves.
4. "What's your change-request process and rate?" Scope will change. The question is whether the process is defined or improvised.
5. "What does month 13 cost me?" If they cannot answer with a number, they have not built something they expected to maintain.

And one for yourself: what decision does this app let me make, and what's the cheapest thing that would let me make it? More budgets have been saved by that question than by any negotiation.

FAQ

Can I really build an app for under $10,000?

Yes, with real constraints. A no-code build or a well-briefed freelancer can produce a working single-platform app with one core workflow for $3,000–$10,000. What you cannot get at that price is custom design, meaningful QA, a scalable backend, or anyone available in six months. For validating an idea, that is often a fair trade. For running a business on it, it is not.

Why do agencies quote so much more than freelancers for the same brief?

Because they are quoting different things. An agency price includes design, QA, project management, a delivery process, and a replacement if someone leaves — plus their sales cost, bench time and margin. A substantial share of an agency's price — in our own business, roughly half — is not the developer's hands on the keyboard. Sometimes that overhead is exactly what you are buying. Sometimes it is not, and the freelancer is the better call.

How much should I budget for the second year?

20% of build cost if you plan to change nothing, 30–40% if you plan to keep improving it. Both are higher than the commonly quoted 15%, because the 15% floor covers keeping the lights on and excludes the feature work any live product needs. Budget the second year before you commit to the first.

Is it cheaper to build one app for both platforms?

Almost always, yes — cross-platform frameworks like React Native and Flutter avoid duplicating the UI layer, and vendor comparisons put the saving at 30–40% versus two native builds, though those comparisons are published by vendors who sell cross-platform work. The exceptions are apps whose value is in graphics, AR, sustained background processing, or heavy camera work. If you cannot name the specific native capability you need, you do not need native.

How do I stop the project from overrunning?

You mostly cannot eliminate it, so plan for it. Flyvbjerg and Budzier's 1,471-project dataset found a 27% average overrun with a fat tail — one in six projects overran by 200% on average. The practical moves: hold a 20–30% contingency you do not tell the vendor about, fix the deadline and flex the scope rather than the reverse, ship something to real users inside 90 days, and treat any estimate given before a written spec as a conversation, not a number.

Does AI tooling make this cheaper in 2026?

The honest answer is that the evidence is thinner than the marketing. METR's 2025 randomized trial famously found experienced developers were 19% slower with AI tools on real tasks in familiar repos, while believing they had been 20% faster. METR has since superseded that result: its February 2026 follow-up estimated an 18% speedup for returning developers and 4% for new recruits, but METR itself calls that data an unreliable signal because of severe self-selection, and is redesigning the study. So the direction has probably reversed; the magnitude is genuinely unknown. Where costs have plainly fallen is scaffolding, boilerplate, test generation and documentation. Where they have not fallen at all: requirements, architecture, integration with systems you do not control, QA on real devices, and the decade of maintenance afterward. Those are most of the bill.

Sources

• U.S. Bureau of Labor Statistics — Software Developers, Occupational Outlook Handbook (median $133,080, May 2024)
• Flyvbjerg & Budzier, "Why Your IT Project May Be Riskier Than You Think," Harvard Business Review, September 2011 (1,471 projects; 27% average cost overrun; one in six overran 200% on cost, ~70% on schedule)
• Flyvbjerg, Budzier, Aaen, Keil & Zottoli, "The Uniqueness of IT Cost Risk: A Cross-Group Comparison of 23 Project Types," Project Management Journal (published online 7 July 2025; 2026 issue)
• METR, "Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity," July 2025 · arXiv:2507.09089 · METR, "We are Changing our Developer Productivity Experiment Design," February 2026
• GoodFirms — How Much Does It Cost to Develop an App (developer survey) (simple app 3–7 weeks; medium complexity $61,000–$69,000)
• Clutch — App Development Pricing Guide (page updated 5 August 2026; typical $10,000–$49,999, average $90,780)
• Second Talent — Cost to Hire a Mobile App Developer by Country (updated May 2026) · Uvik — Offshore Developer Rates by Country (April 2026) · Arc — Freelance Mobile Developer Rates (undated)
• Cleveroad — App Maintenance Cost (cites CISQ, 15–20%/yr) · Savi — Software Maintenance Costs (15–20% rule; IEEE 60/40 lifetime split)
• Google Play Console Help — App testing requirements for new personal developer accounts
• Google Play Console Help — Update regarding Google Play's policies for developers serving users in the US
• RevenueCat — The 15% App Store Fee: A Guide for Developers
• Level Access — European Accessibility Act compliance overview
• Leaper — Supabase vs Firebase cost breakdown, January 2026
• No Code MBA — FlutterFlow Pricing (updated March 2026)
• Capgemini / Sogeti — World Quality Report 2025–26 (registration required)
• Chop Dawg — Native vs. Cross-Platform in 2026 (vendor-published; directional only)

Prices in this guide are US dollars and reflect market rates observed through mid-2026. Figures identified as our own estimates are practitioner judgement, not audited data. Rate benchmarks from agency and directory sources are self-reported and should be treated as directional. App store commission rules are under active litigation in the United States and may change; verify before pricing. Next scheduled review: February 2027.

Related

• App development — prices and dated delivery
• How much does a product launch video cost in 2026?
• How much should a landing page cost in 2026?

Related: Where to submit AI music for review in 2026 — ten platforms compared on price, payment and AI policy.