App Development

How to build an app for your business without wasting $50k

Validate before you build, cut to one job, test with five users, ship in two-week slices. A worked example, plus the questions to ask a developer.

Duraan AliUpdated 17 min read
Three people sticking hand-drawn sketches of phone screens onto a whiteboard in a bright studio

Most of the money wasted on business apps is spent before anyone writes a line of code. It goes into features nobody asked for, a scope nobody tested, and a launch date nobody planned slack around. The process that avoids it is unglamorous: prove demand with something fake, cut the app to a single job, test a clickable prototype with five real people, then build in two-week slices you could stop at any point without losing your money.

Rather than describe that process in the abstract, I am going to run one organization through it. The organization is hypothetical. It is not a client, it is not based on a client, and none of the numbers in the scenario are anyone’s real numbers. I will call it Maple Tutoring: a made-up volunteer tutoring nonprofit in St. Paul with a coordinator, about forty volunteer tutors, and about a hundred and twenty students, running on spreadsheets, group texts, and goodwill. The board has asked for “an app.” Here is how I would walk them from that sentence to something worth paying for, and where I would tell them to stop.

For what the market charges at each step, with every figure attributed, read how much it costs to build an app. This post is about the process, and it quotes only the few figures it has verified itself.

Step 0: what does “an app” mean when the board says it?

Maple’s coordinator arrives with a list. Tutor profiles. A schedule each tutor can see. Session logging with notes. In-app messaging between tutors and the coordinator. A parent portal to see progress. A donate button. A resource library of worksheets. Badges for tutors who hit milestones.

That is eight features and at least three different products. Before anything else, I would ask the coordinator to describe a normal Tuesday. The answer, in this scenario, is that she spends two hours a week chasing tutors by text to find out whether sessions happened and what was covered, because the funder’s report needs session counts and the spreadsheet is always behind.

That Tuesday is the app. Everything else on the list is a nice idea attached to it. Web app vs website vs mobile app explains why the word “app” hides those category differences.

Step 1: how do you know anyone will use it?

Before Maple spends anything, it runs a test that produces a yes or a no. There are three that work for small organizations, and none of them requires a developer.

The spreadsheet test. Put a shared Google Sheet or Airtable base in front of the tutors with one row per session and three columns: date, student, what was covered. Give them a form that writes to it. If forty tutors log their sessions for three weeks without being chased, Maple has learned two things: the tutors will do the work when it takes under a minute, and a form plus a sheet might be the whole answer.

The concierge test. Do the thing by hand for two weeks. The coordinator texts every tutor after their scheduled slot with the same three questions and types the answers in herself. She will learn what the workflow actually is, which is almost never what she assumed, and she will find out whether tutors reply at all.

The landing page test. This one is for ideas aimed at the public. Build a single page describing the app as though it exists, with a sign-up form, and send it to the people who should want it. If forty people see it and two sign up, you have your answer. Maple’s tutors are a captive group, so this test matters less here, but for a product you intend to sell it is the first thing to do.

Validation method What it costs Time What it proves
Shared spreadsheet plus a form A free or low-cost subscription 1 – 3 weeks Whether you need software at all, or just structure
Concierge (do it by hand) Your own hours 2 – 3 weeks The real workflow, and whether people engage
Landing page plus sign-up A page on your existing site 1 week Whether strangers will give you their email for this

Here is the part that costs me work. If the spreadsheet test solves Maple’s problem, Maple should not hire us. A form, a sheet, and a weekly reminder is a legitimate finished answer, not a stepping stone. The coordinator gets her Tuesday back and the funder gets the session counts. Come back when you hit a wall you can name: tutors who need it to work offline in a church basement, a permissions model the sheet cannot express, or an integration with a student records system.

In this scenario, let us say the test came back mixed. Tutors logged sessions when reminded and stopped when they were not, the sheet filled up with inconsistent student names, and three tutors asked whether they could see their own upcoming schedule in the same place. That is a yes with a shape to it: the thing has to be fast, it has to remind people, and it has to show a tutor their own schedule next to the log.

Step 2: what is the one job this app does?

Write it as a single sentence with a subject and a verb. Not “a platform for our tutors,” which is not a job. For Maple: “A tutor logs a completed session with a note in under a minute, from their phone, right after the session.”

Now score the original eight features against that sentence.

Feature Does the job fail without it? Needed by most users in week one? Verdict
Session logging with notes Yes Yes Version one
A tutor’s own upcoming schedule Yes, tutors asked for it and it tells them what to log Yes Version one
Sign-in and a tutor’s student list Yes, the log has to know who is logging what Yes Version one, structurally required
Coordinator view and export for the funder report Yes, it is the reason the job exists One user, but the one who pays Version one, kept minimal
In-app messaging No, group texts work and nobody asked No Version two, probably never
Parent portal No, and it is a different product with different users No Separate project, if ever
Donate button No, the website already does this No Cut
Resource library No, a shared folder does this No Cut
Badges and milestones No No Cut

Eight features became four screens: sign-in, my schedule, log this session, and a coordinator table with an export button. That is a real product, and it is roughly what every good first version looks like: a sign-in, a list, a detail, an action, a confirmation.

The rule that did the work: if removing a feature does not break the one job, it does not ship in version one. Most lists we are handed shrink by about half under it, and nobody misses the removed half after launch. Say no in writing when you defer a feature, because “not in version one” is a decision, and if it is not recorded it will come back in week ten as a surprise.

Every feature you add is not a one-time cost. It is build time, plus testing surface, plus store review risk, plus a maintenance obligation every year the app exists. That is why feature creep is the second most expensive mistake on the list at the end of this post.

Step 3: why test with five users before writing code?

Because fixing a screen in a prototype takes an hour and fixing it in a shipped app takes a sprint. A clickable prototype looks and behaves like the real app and can be in front of tutors a few weeks after kickoff. It is a defined piece of work with a defined output, which means it is a phase you can get a fixed quote on by itself, before anyone has committed to a build.

Five testers is not a made-up number. Jakob Nielsen’s Nielsen Norman Group article from March 2000, “Why You Only Need to Test with 5 Users,” models each participant as revealing about 31 percent of a design’s usability problems and concludes that the first five find about 85 percent of them. Past five, testers mostly repeat what the first five already told you, and the budget is better spent on a second round after you fix things.

Test with the actual people. For Maple, that is five tutors, not the board and not the coordinator. Hand them a phone with the prototype, give them one task in one sentence (“you just finished with Amina, log it”), and then say nothing while they work. Every time you jump in to explain, you erase the finding.

Watch for where they hesitate, where they tap something that is not a button, and where they ask “wait, what does this do?” Write those down verbatim. In a scenario like Maple’s, the likely finding is that the “log this session” screen asked for too much: tutors want one tap for “it happened,” one optional note, and nothing else. Change the design, then run it again with three more tutors.

Step 4: web app, PWA, cross-platform, or native?

This decision is usually made emotionally and it should be made from the facts of your users. For Maple, the facts are: tutors use their own phones, a mix of iPhone and Android; nobody needs to find this app in a store; it has to work in a church basement with bad signal; and the reminders matter.

That points to a web app built as a progressive web app. It runs in a browser, installs to the home screen, caches the schedule so the log screen opens without signal, and sends reminders. The honest caveat is iPhone. Apple’s WebKit team shipped web push for home screen web apps in iOS 16.4 in February 2023, but it only works after the user adds the app to the home screen from Safari, so Maple’s onboarding has to include that step, in person, at the tutor orientation. For forty tutors who already show up to an orientation, that is a manageable instruction. For a public audience it would be a drop-off.

If Maple later needs a store listing, say because a partner school wants to point parents to it, Capacitor can wrap the same web app into a store-ready shell without a rewrite. If it ever needs background location or real-time camera processing, which a tutoring log does not, that is the point to consider native. Native vs cross-platform covers that decision properly, including the store rules that apply to every approach, and I will not repeat it here.

The question that decides it for most small organizations: does someone need to find this in the App Store, or will you always be the one handing them the link? Internal tools for your own staff and volunteers almost never need a store listing. Customer-facing products with repeat weekly use usually do.

One Minnesota-specific note that is not hypothetical. If you run a home care, PCA, or waiver services agency and the app you have in mind is visit tracking, stop and read the Department of Human Services EVV page first. Minnesota requires electronic visit verification for personal care and many waiver services under the 21st Century Cures Act, provides HHAeXchange as the state system at no cost to providers, and requires any third-party system to connect to it. Paying a developer to build a parallel one is an expensive way to create a reconciliation problem the state already solved.

Step 5: how should the build actually run?

In slices, with working software every two weeks. Not “design phase, then a four-month silence, then a demo.”

A slice is a thin vertical cut through the whole product, from the button a tutor taps down to the database row it writes. For Maple, the plan would look like this.

Slice What works on a real phone at the end of it
Weeks 1 – 2 Sign-in against the real backend; the coordinator can add a tutor and the tutor can log in
Weeks 3 – 4 A tutor sees their own upcoming schedule, loaded from real records the coordinator entered
Weeks 5 – 6 A tutor logs a completed session with an optional note; it saves and shows a confirmation
Weeks 7 – 8 The coordinator table shows every logged session and exports the funder’s report
Weeks 9 – 10 Offline caching of the schedule, home screen install, and reminders, tested on both iPhone and Android

Each row is something you can open on your own phone and press. If you can see working software every two weeks, you can stop the project at any point and still own something. If the first thing you see is at month four, you have handed over most of the money on faith. Maple could stop after week eight with a working log and a funder report, and that would be a success.

Three practical rules for this phase:

  • You need one decision-maker on your side. For Maple that is the coordinator, not the board. Every extra approver adds days per question, and app builds run on dozens of questions.
  • Get test builds on real devices, not a simulator on a developer’s laptop. For a web app that means a staging address the tutors can open. For a store app it means TestFlight on iOS and an internal testing track on Google Play.
  • Instrument three events before launch: finished onboarding, logged a first session, came back in week two. Without them, every decision for six months is a guess.

Step 6: what should you ask a developer before you sign?

These are the questions I would want Maple to ask us, and anyone else they talk to. The questions to ask a web designer apply almost word for word; these are the additions for software.

  1. Will I have a running build on my own device within the first month? If the answer is no, ask why the first two weeks cannot produce a sign-in screen against a real backend.
  2. Can you quote the prototype as its own phase, with a fixed price, before quoting the build? A yes means they are willing to let you walk away with something useful. A no usually means the build price was decided before they understood the problem.
  3. Who owns the code, the app store accounts, and the hosting accounts from day one? The answer should be you, in your name, with the developer as a collaborator. Do you own your website explains why this matters, and it matters more for software.
  4. What happens when iOS and Android ship their fall releases? A real answer names the yearly retesting and dependency work and what it costs; a hand-wave means it will be your problem in eighteen months.
  5. Which of my features would you cut? A developer who cannot name one has not thought about your scope, or is happy to bill for all of it.
  6. Have you shipped to both stores before, and what rejections did you get? Everyone who has shipped has been rejected at least once. The useful answer describes the rejection and the fix.
  7. If an AI tool built the first draft, who on your team can explain every line of it? Fast code is fine. Code nobody understands is a liability you inherit. Our post on fixing a vibe-coded app covers what that liability looks like from the inside.

On the contract: fixed price works for a defined phase, because discovery, a prototype, and a design system have defined outputs. It works badly for a full build on vague requirements, because the developer prices in risk and then spends the project defending the scope line. Fix the price on the early phases, then price the build from what they produced. That is how TayoPro scopes software: a fixed price per phase, in writing, and you can stop after any of them.

Step 7: what happens at app store submission?

Maple’s PWA skips this entirely, which is one of the quiet arguments for it. If your project does go to the stores, this is where first-time app owners lose weeks they did not plan for, and none of it depends on which framework you chose. The facts below are from the store pages, read in November 2026.

Apple’s Developer Program is a $99 annual membership, and a Google Play developer account is a one-time $25 registration fee. Neither store publishes your app because you paid. Apple’s App Review page says it typically reviews at least 50 percent of submissions in under 24 hours and 90 percent in under 48, which sounds fast until a rejection resets the clock.

The rejections that hit small-organization apps most often, from Apple’s App Store Review Guidelines:

  • Guideline 4.2, minimum functionality. Your app should include features, content, and UI that elevate it beyond a repackaged website. This is the single most common rejection for small business apps, and it is the reason “just turn our site into an app” is not a project.
  • Guideline 5.1.1, account deletion. If your app supports account creation, you must also offer account deletion within the app. Not by emailing you.
  • Guideline 3.1.1, in-app purchase. Unlocking features or content inside the app, including subscriptions and premium content, must go through Apple’s in-app purchase, with Apple’s commission attached. Physical goods and real-world services are treated differently.
  • Broken or incomplete builds. Placeholder text, a crash on one device, or a login the reviewer cannot get past because nobody supplied a demo account.

Google Play’s gate is different. Its testing requirements page says a personal developer account created after November 13, 2023 must run a closed test with at least 12 testers opted in continuously for 14 consecutive days before applying for production access. People find this out the week they wanted to ship. Both stores also require privacy disclosures, Apple as privacy labels and Google as the data safety section, and both must accurately describe what you collect. Fill them out with your developer, do not guess, and do not let anyone check boxes to make the form go away.

Plan two to four weeks of slack between “the app is done” and “the app is public,” and do not book a launch event before you are approved.

Step 8: launch is where the cost starts

The app going live is the beginning of the spending, not the end of it. Chopdawg’s July 2026 guide to app maintenance and support costs in the USA puts yearly maintenance at 15 to 20 percent of the original build cost, and Space-O Technologies’ maintenance cost guide, updated October 21, 2026, gives the same 15 to 20 percent. Both are software vendors, so treat the figure as a planning rule rather than a measurement, but it is the rule the industry plans with, and the 2026 guides collected on our pricing page run as high as 25 percent.

What it buys: operating systems update every fall and break things. Frameworks and dependencies need patching. Google Play’s target API policy, read in November 2026, requires new apps and updates to target Android 16 from August 31, 2026, and the bar rises every year. Apple periodically requires builds against a recent SDK to stay in the store. Servers, databases, push services, and error monitoring all have monthly bills. And people will find bugs you did not.

Skip two years of this and you are not maintaining an app anymore, you are planning a rebuild. Maple’s PWA carries less of it, since there is no store and one codebase, but the backend, the dependencies, and the fall retesting still exist. Budget for them the way you would budget a website care plan, as a line that recurs.

What are the five most expensive mistakes?

Building before validating, because a finished feature nobody opens is the most expensive line in app development. Feature creep, because every “while we are in there” adds build cost now and maintenance cost forever. A fixed-price contract on unclear scope, which turns the project into a negotiation about the scope line. No analytics in version one, which means six months of guessing. And no maintenance budget, which turns an asset into a write-off eighteen months after launch. Four of the five are decided before development starts, which is why the steps above spend so long there.

When should you not build an app at all?

If people interact with your organization a few times a year, you do not need an app, you need a fast website. Nobody installs software to check your hours, read a program description, or make one donation. A well-built nonprofit or small business site does those jobs better and does not require anyone to download anything. Maple’s donate button belongs on the website, which is where it already was. If the real problem is internal, look at a web app or internal tool before a mobile app; it skips both stores, updates instantly, and works on a laptop. And if your website is slow or invisible on Google, fix that first. Website or app has five tests for the decision, and why your website is slow covers the fix.

What to do next

Pick the validation test that fits your situation and run it this week. If it comes back no, you have saved the budget. If it comes back yes with a shape, write the one-sentence job, score your feature list against it, and get the prototype quoted as its own phase before anyone quotes a build.

If you want a second pair of eyes on any of that, tell us what you are working on. TayoPro builds web applications and mobile apps in fixed-price phases you can stop after, and we design and build websites and keep them maintained month to month, all at a fixed price quoted in writing. We tell a fair number of people that a website or a spreadsheet is the better answer, and we will tell you plainly if you do not need any of it.

FAQ

Questions people ask about this

How do I validate an app idea without spending money?

Run the manual version yourself for two weeks. Take the requests by phone, text, or a shared spreadsheet and do by hand what the app would automate. If fewer than ten people engage, or you can keep up manually without strain, you do not have an app problem yet. This costs nothing and it has stopped more bad builds than any other step.

How many features should the first version have?

One job, done completely, plus whatever that job structurally requires. In practice that is three to six screens: a sign-in, a list, a detail view, an action, and a confirmation. Every additional feature adds build time, testing surface, store review risk, and a maintenance obligation for as long as the app exists. You can add features after launch with real usage data. You cannot get back the money spent building the wrong ones.

Should I sign a fixed-price contract for an app?

Fixed price works for a defined phase, such as discovery, a prototype, or a design system, because those have defined outputs. It works badly for a full build on vague requirements, because the developer prices in risk and then defends the scope line instead of solving your problem. Fix the price on discovery and the prototype, then price the build from what those phases produced. That is how TayoPro scopes software: a fixed price per phase, in writing, and you can stop after any of them.

How long does app store review take?

Apple's App Review page, read in November 2026, says it typically reviews at least 50 percent of submissions in under 24 hours and 90 percent in under 48. That clock resets on a rejection, and the common rejections for small-organization apps are a repackaged website, no in-app account deletion, and missing privacy disclosures. A new personal Google Play account must also run a closed test with 12 testers for 14 consecutive days before production. Plan two to four weeks of slack, not two days.

What analytics do I need on day one?

Enough to answer three questions: how many people finished onboarding, how many completed the core job at least once, and how many came back in week two. Three events, instrumented before launch. Launching without them means every decision for the next six months is a guess, and most product analytics tools have a free tier that covers a small app.

Can I build a business app myself with no-code tools?

Often yes, and you should try first. A spreadsheet database with a no-code app layer handles scheduling, checklists, intake, and inventory for many small organizations. Hire a developer when you hit a wall you can name: offline use, complex permissions, an integration with a billing or records system, or a public store listing. If you already built something with an AI tool and it is half-working, our vibe-coded app post covers how to get it finished safely.

Keep reading

Start a project

Ready to take your business to the next level?

Let's work together to create solutions that make a real impact. Tell us about your project and we'll come back with a clear plan.

(612) 293-0636info@tayopro.com