How much does it cost to build an app?
The short answer is $15,000 to $90,000 for most projects. The useful answer is which end you’re on, and why.
- $15K – $90K
- What most custom applications cost
- $9K – $15K
- The admin application, the most underestimated line item
- 10–13%
- QA and release, as a share of everything there is to test
Numbers first, reasoning after. The answer does depend on a handful of things, and this page names them, but you can have the ranges before you read a word of explanation.
These are ranges for custom software built by a professional team: design, engineering, testing, and someone accountable for the schedule. A template costs less. So does an offshore hourly rate multiplied by an optimistic guess, right up until you compare what arrives.
The short answer
Three kinds of project cover most of what gets built. Find the description that sounds most like yours.
$15,000 – $25,000
A focused first version
About 9–15 weeks
One thing done properly, for one kind of user, with a new backend behind it. Enough to put a real product in front of real people and learn something you couldn’t have guessed.
$30,000 – $55,000
A production application
About 3–5 months
More than one job to do, more than one kind of person doing it, and somewhere for whoever runs the place to keep an eye on it. Most software that gets used daily lands here.
$55,000 – $90,000
A complex build
About 4–6 months
Works offline, talks to hardware, integrates with a system you already run, or handles regulated data. Any one of these puts you here. Two of them and you start at the top of the range.
These are the three kinds of project that show up most often. The lines between them are soft. Plenty of work lands in the gaps, and a project can be complicated in one way and straightforward in every other. The calculator prices what you describe instead of sorting you into a box.
Why the range is that wide
The width comes from the phrase itself. "An app" describes a five-screen utility and a warehouse system used by sixteen thousand people, and nothing in those two words separates them.
Two apps with the same number of screens can require wildly different engineering. Take a screen that shows today’s jobs. Version one fetches a list from an endpoint and draws it. Version two has to work in an elevator, hold the list on the device, let the user complete a job while disconnected, then reconcile the result against changes a dispatcher made to the same record from an office. The wireframes are indistinguishable. One is two weeks of work and the other is two months.
This is why a single confident number deserves suspicion. A figure with no range attached tells you how much the estimator was willing to say out loud, and nothing about how much is still unknown.
What you’re actually paying for
You’re buying somebody’s time, and how it gets packaged varies. Some shops quote a fixed price for a defined scope, some bill as they go, some do a bit of both. None of those is a trick. What matters is that you and they mean the same thing by the words in the brief, because that’s where the money actually moves.
- Two numbers can be far apart and both be right
- Two proposals $20,000 apart usually aren’t pricing the same project. One assumed you had an API, the other assumed building one. One counted the office side, the other read past it. That gap is yours to close, and closing it’s faster than shopping further.
- Say what you already have
- A working API, finished designs, a brand, a data set that has to come across. Each one takes real work off the number, and none of them get assumed in your favor if you don’t say so. Being specific about what exists is the cheapest thing you can do to a quote.
- Two of the eight are shares, not sums
- QA and release runs at 10–13% of everything there is to test, because more built means more to test. Project management runs at 7–10% of the whole. Both scale with the build, which is why projects of the same size land near the same share.
- 10–13%
- QA and release, as a share of everything there is to test
- 7–10%
- Project management, as a share of the whole
Those hours divide across eight kinds of work, and knowing the eight is how you tell a thorough plan from a thin one.
Discovery and architecture: turning the idea into a scoped, buildable plan.
Product and UI design: screens, states, and a design system, reviewed before anyone writes production code.
Application development: the app itself, meaning screens, navigation, state, device behavior, and release builds.
Backend and data: APIs, the data model, storage, authentication, environments, deployment.
Integrations: connecting to systems you already run, and handling it when they misbehave.
The admin application: the office-facing side, which is effectively a second product.
QA and release: device testing, edge cases, store submission, and the review cycles after it.
Project management: planning, coordination, demos, and keeping scope and schedule honest.
The last two are shares, not fixed figures: QA runs at 10–13% of everything there is to test, project management at 7–10% of the whole. A proposal missing both is about 20% short before anyone has argued about a feature.
The four questions that move the number most
Answer these before you speak to anybody and you’ll get quotes worth comparing, because everyone will be pricing the same project. What a studio needs to know to quote you puts a dollar figure on each one, measured on a single build.
- 1. Do you already have a backend?
- A finished, documented API takes the server-side work down to roughly a third of building one from scratch. A partial one, which is the usual answer, helps much less, because the missing part tends to be the part your app needs. Be specific: "we have an API" and "we have a database and an internal tool that reads it" are very different answers. What a backend is.
- 2. How many different kinds of user are there?
- Not how many users. How many kinds. One is the baseline; the second one costs money, because it brings screens, rules, and tests of its own. Four kinds of user is a much larger project than two, and it’s the most common source of quiet scope growth after a project starts. What a user role is.
- 3. Does it have to work without a network, or talk to hardware?
- Either of these moves a project up a band on its own. Offline means local storage plus synchronization plus conflict resolution plus testing on bad networks. Hardware means pairing, permissions, protocol work, and a test pass on real devices in the states they fail in. What offline mode is.
- 4. Is regulated data involved?
- Health records, card data, government work, or a formal accessibility conformance requirement. Each of these changes how the software gets built, not just what gets documented at the end, adds somewhere between 5% and 35% depending on the regime, and turns into a rebuild if you retrofit it later.
What the very cheap number leaves out
A quote dramatically below these ranges is usually a smaller scope wearing the same words. The gap is normally made of the same four things.
The admin application, described in one sentence and priced at zero, then rediscovered as "phase two" at close to the cost of the app itself.
QA beyond the developer checking their own work, which finds a fraction of what your users will.
Project management, which stops being billed long before it stops being needed. Then you do it, for free, in the evenings.
Discovery, cut on the grounds that the requirements are obvious. They’re obvious right up until two people who both understood them perfectly turn out to have understood them differently.
The way to compare quotes fairly is to give everyone the same brief. Write down what already exists, how many kinds of user there are, and whether the office side is in scope. Numbers that came from the same facts can actually be compared.
How long it takes
A focused first version is usually 9–15 weeks. A production build is 3–5 months, and a complex one 4–6. Those are elapsed calendar weeks with a real team on it, instead of hours divided by an idealized working day.
- 2–3 weeks
- Fixed at the ends: kickoff, your review cycles, and store review
- 8 weeks
- The floor, whatever the arithmetic says
Two to three of those weeks are fixed no matter how the project is staffed: kickoff and discovery, your own review cycles, and app store review at the end. Adding engineers compresses the middle of a project and does nothing to the ends. A two-week wait for a decision on your side costs two weeks whether one person or six are waiting for it, which is why the fastest projects are usually the ones with a single available decision-maker, not the ones with the biggest team.
What happens after you launch
Spending continues after launch, and a plan that assumes otherwise gets surprised around month eight. Apple and Google ship operating systems every year. Dependencies and SDKs go out of support. Store policies change with deadlines attached. Bugs appear that only exist at real volume, and real users immediately suggest changes nobody thought of.
- 15–20%
- Of build cost per year. A planning convention the calculator does not produce
A common planning figure is 15–20% of the build cost per year for an application in active use, less if it’s stable and nobody is asking for anything new. That figure is a convention, not something this calculator computes. the estimate here covers building the thing, not running it. Put a number in the slot before you commit to the build.
Common questions
How much does it cost to build an app?
Most custom applications cost between $15,000 and $90,000. A focused first version with one workflow and one type of user is usually $15,000–$25,000. A production application is usually $30,000–$55,000: a few things people do every day, more than one kind of person on it, and an office side behind it. Offline operation, hardware, integrations with existing systems, or regulated data push a project to $55,000–$90,000 and sometimes beyond.
Can I build an app for $5,000?
Not with a backend, two platforms, and someone answering the phone when it breaks. What $5,000 buys is a clickable prototype, which is a real purchase for testing an idea as long as you’re buying it knowingly as a prototype and not as a discounted app. What $10,000 buys covers the other things sold at that price and where the floor under a custom build actually sits.
How much does it cost to maintain an app?
A common planning figure is 15–20% of the build cost per year, and it covers keeping the application working: annual OS releases on both platforms, dependency and SDK updates, store policy changes, and defects that only appear at real volume. New features are a separate budget, and what app maintenance actually costs prices both lines with the engine. An app nobody touches for eighteen months is drifting, and the bill for that arrives all at once.
Should I get a fixed price or pay hourly?
A fixed price is the same price with the risk transferred and a margin added for carrying it. It works well when the scope is settled and both sides can describe exactly what "finished" means. It works badly when the scope is still moving, because every change becomes a negotiation and the incentive quietly shifts from building the right thing to building the agreed thing.
Why is this calculator free?
It was built by a team who design and build custom software for a living. A conversation with someone holding a realistic budget is better than one with someone holding a fantasy, whether or not that conversation happens with us. The estimate itself is free and complete, with nothing held back for a follow-up call. Your name and work email at the last step are the price of it, and they come to us along with your project description.