What is MVP in app development? Twenty answers from idea to app store

Five tools document all three as of 19 September 2026: FlutterFlow, Thunkable, Expo with EAS Submit, Codemagic and fastlane; Newly covers both stores, but its documented backends are Liquid Backend and Supabase, with Firebase appearing only as the Android push credential.

All checked on 19 September 2026. FlutterFlow connects a project to Firebase for auth, Firestore and storage, deploys to App Store Connect with an API key, and to Google Play with a service account after a first manual AAB upload. Thunkable wires a Firebase Realtime Database and sign-in, sends the iOS build to App Store Connect with an app-specific password, and emails you an AAB in about 30 minutes to upload to Play Console yourself.

Expo’s EAS Submit uploads to App Store Connect and Google Play, and Expo documents Firebase through the JS SDK or React Native Firebase. Codemagic and fastlane are release pipelines, not builders: both publish to both stores, and both reach Firebase through App Distribution, which ships test builds.

Newly builds the iOS app and submits it to App Store Connect automatically, and produces a signed AAB plus a Play Console connection for Android. Its data layer is Liquid Backend or Supabase; Firebase appears only as the FCM credentials for Android push. If Firebase is a hard requirement, pick from the five above; if not, our guide to choosing a backend for a mobile app explains why we would not start a new app on it.

Four steps: write who has the problem and when it bites in one sentence, watch five people who have it cope today, name the single assumption that would kill the idea if it were wrong, and build only the screen that tests it, on a real phone, within a week.

The order matters more than the steps, and most people run it backwards: they start at step four with the whole app, discover at step two that the problem is not what they thought, and never write step one at all. A sentence you could text to a friend is the cheapest prototype there is, and if it will not fit in a text it is not one idea yet.

Step two is watching, not pitching. People will say an idea sounds great because that costs them nothing; what they do today when the problem bites is the only honest data you can get for free. Step three is where most ideas die, and that is the point of doing it early, while dying is cheap. Step four is the first thing that costs money, and on an AI builder it costs a prompt.

The full route, including how the validation step feeds the first build and the store submission, is our guide to developing an app idea from concept to launch.

Yes, the question underneath is which mobile development software integrates with Firebase, App Store Connect and Google Play, and as of 19 September 2026 the tools whose own docs cover all three are FlutterFlow, Thunkable, Expo EAS, Codemagic and fastlane.

The phrasing is a search assistant’s artefact, not a different question, so the tool-by-tool answer with each vendor’s documentation is two questions up. Newly sits beside those five for both stores but on a different backend, which is a decision our guide to choosing a backend walks through.

Yes: one of us built a five-screen employee management app from one 268-character prompt in one sitting on 10 September 2026, and Google's developer blog records that Alarmy, now past 90 million downloads, began as a basic version one computer science student built and showed at a tech event.

The Alarmy story is on Google’s own Android Developers Blog, in a #WeArePlay post dated 6 November 2025: a student who could not wake up for morning lectures built a basic version, showed it at a tech event, and the prototype became the app. Ours, smaller: one 268-character prompt, 47 actions and 118 seconds of model thinking on the frontend pass, and a three-minute connection stall that a page reload cured (newly.app/build/employee-management-app, 10 September 2026).

The realistic scope is what makes the yes true: one job, one kind of user, three to five screens, one place the data lives, and no payments or roles in version one. One person absorbs a wrong guess and a crash in an afternoon, not two roles, offline sync and a subscription flow, and the tool will build all three if the prompt leaves the door open.

The one thing a solo developer cannot do alone is the Android gate: a personal Google Play account created after 13 November 2023 must have 12 testers opted in for 14 consecutive days before production access. Recruit them early. What to cut so one person can carry the rest is our MVP app development guide.

Account setup and review, not the build: Apple says 90% of submissions are reviewed in under 24 hours, Newly's docs put a first review at 1 to 3 days, and a new personal Google Play account cannot release until it has run a 14-day closed test with 12 testers, all read on 19 September 2026.

The build is now the short part. Once your Apple account is connected, Newly auto-provisions the signing keys, compiles the iOS app in typically 5 to 15 minutes and submits it to App Store Connect without Xcode. Apple’s own figure is on its App Review page: on average, 90% of submissions are reviewed in less than 24 hours. Newly’s guide, written for first-time accounts, says 1 to 3 days.

Before either clock starts you need the accounts. Apple says to contact them if membership confirmation has not arrived within 24 hours of purchase; Google’s identity check is usually a day or two by Newly’s Play setup guide. Then the listing: screenshots per device size, the App Privacy questionnaire, an age rating, and a working demo account if there is a login. Missing demo credentials is one of the most common rejection reasons, and each rejection costs a full review cycle.

So open the accounts first, in parallel with the build, and start the Play closed test the day you have an AAB. The order, step by step, is our guide to publishing an app to the App Store.

124 USD in fees before anything else, 99 USD a year for the Apple Developer Program and a one-time 25 USD for a Google Play developer account, both read from Apple's and Google's own pages on 19 September 2026, plus whatever builds the app, which on Newly starts at $25 a month for 50 credits.

Apple’s enrollment page states the Apple Developer Program annual fee as 99 USD, in local currency where available. Google’s Play Console help states a US$25 one-time registration fee. Neither can be waived. Apple’s renews every year whether or not you ship an update; Google’s does not.

The build cost depends on the route. Newly’s Agent plan is $25 a month for 50 credits (newly.app/pricing, read 19 September 2026); one Build prompt is one credit, Plan and Ask prompts are a tenth of one, and the five-screen app we built in September took two Build prompts. Hiring is a different order of magnitude, and our app development cost guide puts agency and freelance ranges beside the subscriptions, so we will not repeat them here.

Two costs the fee pages do not mention: both stores require a public privacy policy URL, so budget a page for it, and you need a real phone to test on, because the preview gets layout right and lies about keyboards and scrolling.

Scope it to one job for one kind of user, write that job as a prompt of two or three sentences, and build the screen that does the job before anything that supports it; the prompt below did exactly that on 10 September 2026 and produced five screens in one pass.

The prompt is the scope. Two or three sentences that name who uses the app, the one job it does and what they see first; leave out roles, payments and settings, because those are version two and an AI builder will invent them, confidently, if the prompt leaves a gap. The one below is the exact brief that built five screens on 10 September 2026, and the one thing wrong with it is that it never named the industry, so the tool guessed a coffee shop.

Build an employee management app for a small team. Managers see a roster of employees, open an employee to see their details and shifts, assign shifts for the week, and approve or decline time-off requests. Employees see only their own shifts and can request time off.

After the first pass, correct in the words a user would say, one change per prompt, and look at the preview before prompting again. The habit that turns a first build into an MVP rather than a demo is putting it on a phone and in front of a stranger the same day, which is the discipline our rapid prototyping guide is about.

It is the same thing the first answer on this page defines: the version of your app that teaches you the most about customers for the least effort, in Eric Ries's words from 2009, and not the smallest app you can ship.

Ries’s post is linked from the first answer. The reason the word gets asked five different ways on this page is that it is used two different ways: as a stage, the first release, and as a method, releasing in order to learn. The stage is what most people mean; the method is what Ries meant, and it is the useful one, because it tells you when you are done: when you have the answer, not when you have the features.

Three checks tell you whether what you have is an MVP or just a small app. You can say in one sentence which assumption it tests. A stranger can use it without you standing beside them. It measures one number you agreed on before you built it. Miss any of the three and you have a prototype, which is fine, but it is a different thing and it teaches you less.

What goes in and what stays out is our MVP app development guide.

MVP app development is building a mobile app in the order that tests your riskiest assumption first: ship the smallest version a real user can try, measure what they do, and let that decide version two.

It is a sequencing discipline, not a price tier. The order is: the assumption you are least sure of, the smallest screen that tests it, a real user, a number, then the next assumption. Nothing about that order says cheap; it says early. A team with a large budget can run it, and usually should, because the expensive mistake is not an expensive app but a finished one nobody wanted.

For a mobile app in 2026 it means three things in practice. The first build is one prompt or one screen, not a spec. It goes on a phone before it goes in a slide deck, because the preview lies about keyboards and scrolling. And the store submission is part of the loop, not the end of it: a first Apple review is typically 1 to 3 days and a new personal Play account needs a 14-day closed test, so you plan the second version while the first is in the queue.

The costs are further down this page, and the features are in our MVP app development guide.

Turn the idea into one screen a stranger can use, then take it through four gates in order: a build on a real phone, testers, the two developer accounts at 99 USD a year and 25 USD once, and store review, where Apple clears 90% of submissions in under 24 hours by its own figure.

Launching is what happens after the idea has become one screen a stranger can use, and each gate has a number attached. A build on a real phone is minutes on an AI builder: a signed Android APK took about six minutes on 10 September 2026 and installs without any store account. Testers are a gate, not a suggestion, on Android: 12 for 14 consecutive days on a new personal account. The accounts are 99 USD a year and 25 USD once. Review is 90% under 24 hours on Apple’s figure, and up to seven days for Google to grant a new account production access.

The mistake is doing them in the wrong order: idea straight to the accounts, then a build, then no testers when the 14-day clock should already be running. Open the accounts and recruit the testers the week you have a build. The full order, validation included, is our guide to launching an app idea.

Describe the app in plain English to an AI builder such as Newly or Bolt and answer the one question it asks back; Newly's quickstart puts a first build at about 1 to 2 minutes, and our own 10 September 2026 session reached a five-screen preview after 47 actions and 118 seconds of model thinking, though nobody timed it.

Plain English means a paragraph, not a spec: who uses it, the one job, what the first screen shows, what happens when they act. The builder writes the screens and the navigation, compiles a real React Native app and runs it in a phone-shaped preview you can tap. Newly’s quickstart says the first build takes about 1 to 2 minutes; ours also hit a three-minute connection stall that a page reload cured, so plan for five minutes and be pleased.

Expect one question back before it builds, and answer it carefully. On 10 September Newly asked whether to connect Supabase for real logins or build with mock data first, and mock data is why the app arrived with eight invented employees. That question decides whether you get a demo to look at or an app to use.

Bolt does the same job on Expo and hands you the EAS commands to publish; Lovable’s own docs describe it as a platform for web applications, so it is the wrong tool for a phone prototype. Where a prototype ends and an MVP starts is our guide to AI-powered prototyping.

About a week of calendar time, most of it waiting: build with an AI builder that submits for you, Newly compiles the iOS app in typically 5 to 15 minutes and sends it to App Store Connect automatically by its docs on 19 September 2026, then Apple review, which clears 90% of submissions in under 24 hours and a first app typically in 1 to 3 days.

The sequence, with numbers. Enrol in the Apple Developer Program: a few hours to a couple of days by Newly’s guide, and 99 USD. Build with an AI builder and let it submit: typically 5 to 15 minutes on Newly. Install it through TestFlight on a real iPhone and fix what you find; a crash caught there is free, one caught in review costs days. Fill in the listing. Submit. Apple’s figure is 90% of submissions in under 24 hours; a first app typically 1 to 3 days.

Android is slower for a new personal account, not because of the build but because Google requires a closed test with 12 testers for 14 consecutive days before production access, then reviews the application in usually seven days or less. So the honest fastest route is iOS first, with the Android clock running in parallel.

None of it is faster if the app crashes in review, which is why the testing question below matters more than tooling. The step-by-step is our guide to publishing an app to the App Store.

Make one screen that does the job, put it on a phone, and hand it to five people who have the problem; the fuller recipe, with the prompt we ran on 10 September 2026, is under the build question above.

What changes between making an MVP and making an app is what you leave out, so here is the leaving-out list. No accounts until a stranger has used it without one. No payments until someone has asked to pay. No settings screen, no dark mode, no onboarding. One data source, and mock data is acceptable for the first week as long as you know it is mock. If the builder asks whether you want real logins or demo data, demo data gets you screens faster and real logins get you an app you can actually hand to someone; pick on purpose.

The prompt and the first-pass result are under the build question above. What to test before you show it to anyone is further down, and how to decide what a first version includes is our MVP app development guide.

MVP stands for minimum viable product: minimum for the least you can build, viable for the fact that a real customer can get the value from it, and product for the fact that it is a real thing rather than a mock-up; the definition most people quote is Eric Ries's, from 3 August 2009.

Ries’s post of 3 August 2009 is where the phrase acquired the meaning it has now: that version of a new product which allows a team to collect the maximum amount of validated learning about customers with the least effort. Notice that his sentence does not contain the word small. Minimum modifies effort, not ambition.

The three words get misread in three predictable ways. Minimum gets read as cheap, so people ship something broken and call it lean. Viable gets read as working, when it means a customer can get the value; a working app nobody can find the value in is not viable. Product gets skipped, so a survey or a slide deck gets called an MVP when it is research, which is useful but a different thing.

The full definition and what it means for an app is the first answer on this page; what to build first is our MVP app development guide.

Seven things, on a real phone rather than the preview: sign-up and login, the main flow end to end, payments if you have them, offline and slow-network behaviour, every permission prompt, a fresh install on a second device, and the demo account you will hand the reviewer.

1. 1Sign up, log in, log out and back in, on a device that has never seen the app.
2. 2The one main flow, start to finish, with your own data rather than a builder's invented rows.
3. 3Payments, if you have them, in both stores' sandboxes, including a cancelled purchase.
4. 4Airplane mode and a slow connection: what the app shows when the network goes.
5. 5Every permission prompt, and the sentence that explains why you need it.
6. 6A fresh install on a second phone, ideally the other platform.
7. 7The reviewer's path: the demo account works, the privacy policy URL loads, nothing says placeholder.

Real device, not the preview: Newly’s own submission guide opens with it, because a phone has a keyboard, a permission dialog and a flaky network, and the preview has none of them. Item seven costs the most when skipped: a missing demo account is one of the most common rejection reasons. On Android, testers are a gate: 12 for 14 consecutive days on a new personal account.

How to recruit testers, which tools to use and how many rounds to run is our pre-launch app testing checklist.

Enrol in both developer programs, 99 USD a year for Apple and 25 USD once for Google, build a release binary, put it through TestFlight or a Play closed test, fill in the store listing, and submit for review; end to end that usually takes a few days, and most of it is waiting on Apple and Google rather than work on your side.

That sentence is the whole route, and each clause has a number. The accounts are the only money the stores take: 99 USD a year and 25 USD once. The binary is an .ipa for Apple, which Newly builds and submits in typically 5 to 15 minutes, and a signed .aab for Google, which you upload to Play Console yourself. TestFlight needs the Apple program; a Play closed test on a new personal account needs 12 testers for 14 consecutive days. The listing is separate from the build on both stores and takes longer than people expect: screenshots, privacy answers, an age rating, a demo account.

Newly’s guide says to set aside time because getting the accounts approved and the first build through review usually takes a few days end to end, and that most of it is waiting. We agree, and the order that avoids waiting twice is our guide to publishing an app to the App Store.

Yes, within limits: a plain-English description to an AI builder produces a running app with screens, navigation and data you can tap through in a phone-shaped preview, about 1 to 2 minutes for a first build by Newly's quickstart, and it stops at the things only you can decide, such as accounts, payments and the two store listings.

What describing produces is a real app, not a picture of one: a named project, the screens your sentences imply, navigation between them, a data layer, and either invented data or an empty first-run state depending on how you answer the one question the builder asks before it starts. Newly’s quickstart puts a first build at about 1 to 2 minutes, and its documented starter prompt is below, verbatim.

Create a task manager app where users can:
- Add tasks with a title and due date
- Mark tasks as complete
- Filter between active and completed tasks
- Swipe to delete tasks
Use a clean, minimal design with a blue color scheme.

Where it stops: it will not decide what the app is for, it will invent whatever you leave unspecified, and it cannot do the parts that need you, the developer accounts, the store listings, the privacy policy, the demo account for the reviewer. It also stops at the preview until you build: a real phone needs an APK or a TestFlight build. How the prompting works, and what a good description contains, is our text-to-app guide.

An MVP in app development is the first version you can put in front of a real user to learn whether the idea holds, defined by Eric Ries in 2009 as the version that yields the maximum validated learning for the least effort; the full definition is the first answer on this page.

For an app specifically, viable has three parts that a web MVP does not always need. It installs, on a real phone, which means a build rather than a preview. A stranger can use it without you: no walkthrough, no seed data you have to explain. And it produces one number you can read the next morning, because the point of the release is the learning, not the release.

That is why we treat the store submission as part of the MVP rather than something that comes after it. The build is minutes; the accounts are 99 USD a year and 25 USD once; the review is 90% under 24 hours on Apple’s figure. The learning starts when a stranger installs it, and nothing before that counts. What goes in the first version is our MVP app development guide.

124 USD in mandatory fees, 99 USD a year to Apple and 25 USD once to Google as of 19 September 2026, plus the cost of building it, which starts at $25 a month on Newly; the breakdown is under the launch cost question above.

The two figures come from Apple’s enrollment page and Google’s Play Console help, and the Newly price from newly.app/pricing, all read on 19 September 2026. What moves the number is the build route, not the stores: an AI builder subscription is tens of dollars a month, a freelancer is thousands, an agency is tens of thousands, and the comparison, with the hidden costs the fee pages leave out, is our app development cost guide.

One thing that does not cost money but is priced like it does: a new personal Google Play account cannot release to production until 12 testers have been opted in for 14 consecutive days, so the cheapest launch is still two weeks long on Android.