Key takeaways
- For almost every small business, a website beats an app: cheaper to build, discoverable on Google, and zero install friction.
- A custom app costs $15,000–$150,000+ to build, plus roughly 15–20% of that every year to maintain. A great website costs a fraction of that.
- The average phone has 80+ apps installed, and the owner uses 9 a day. Your app has to displace one of those 9 — or get deleted.
- Build an app only when customers will open it weekly or more, or you genuinely need the phone's hardware.
- The smart sequence: website first, watch the data, add a PWA or native app only when usage earns it.
The short answer
Build the website. Not the app. Not both. The website.
We say this as a team that makes money fixing AI-built app prototypes — half our rescue work is founders who built the flashiest thing first and validated last. An app-first strategy is the most expensive way to learn what a website could have taught you for a tenth of the price.
That's not ideology. It's arithmetic. Let's do the math the app-first crowd skips.
The download math nobody shows you
Here is the number that kills most small-business apps before a line of code is written: the average smartphone has more than 80 apps installed, and the average person uses 9 of them a day. (Buildfire's 2026 mobile usage stats.)
Read that again. Eighty-plus installed. Nine used. Your app isn't competing with "no app." It's competing for one of nine daily slots against the bank, the maps app, the group chat, and Instagram.
And the retention numbers are brutal. Day-one retention averages around 24% — three-quarters of people who download an app never open it the next day. By day 30, it's 5–6%. A quarter of all apps are opened exactly once, then abandoned. (AppMakers USA's 2026 industry report, compiled from Sensor Tower, Apple, and Google data.)
You're not asking customers to download your app. You're asking them to delete someone else's.
Now ask the honest question: will your customer open your app more than once a month? A restaurant's regulars might open a loyalty app weekly. A plumber's customers will open his app… never. They'll Google "plumber near me" once every three years and call whoever ranks. An app for that business isn't a product. It's a very expensive business card that lives in a drawer.
What each option actually costs
App development quotes are famous for starting with "it depends." That's true, and useless. Here are the real 2026 ranges from this year's agency surveys — Tvisha, Boolean Inc, Decode Agency, and Clutch, whose 2026 pricing survey puts the average agency-built app at $90,780:
| Custom website | Progressive web app | Simple native app | Serious native app | |
|---|---|---|---|---|
| Build cost | $4,000–$18,000+ | Website cost + uplift | $15,000–$50,000 | $50,000–$150,000+ |
| Timeline | 4–8 weeks | 5–10 weeks | 8–14 weeks | 4–7 months |
| Yearly upkeep | Hosting + small updates | Small | 15–20% of build cost | 15–20% of build cost |
| Shipping an update | Minutes | Minutes | 1–3 days (store review) | 1–3 days (store review) |
A few things the table doesn't show. First, "native app" usually means two apps: iPhone and Android are separate builds. Covering both platforms natively adds roughly 30–45% over a single platform (Boolean Inc's 2026 data) — unless you go cross-platform with Flutter or React Native, which cuts the bill by about a third.
Second, the build quote isn't the budget. The industry benchmark for app maintenance is 15–20% of the build cost every year — OS updates, security patches, bug fixes, hosting. A $60,000 app quietly costs you another $9,000–$12,000 a year, forever. Websites need upkeep too, but it's an order of magnitude smaller. (Our Sharp Site builds start at $4,000 — the entire website costs less than one year of maintaining a mid-range app.)
Third, the store takes its cut. Apple charges $99 a year for a developer account; Google Play charges $25 once. Both take 30% of in-app revenue, dropping to 15% for smaller developers under $1M a year. Sell a $10/month subscription inside your app and the store keeps $1.50–$3 of every payment, every month, forever. Sell the same subscription on your website and you keep all of it minus payment processing.
A $60,000 app is really a $60,000 app, plus ~$10,000 a year, plus 15% of your revenue, plus a second build for the other platform. Quote the whole thing or don't quote at all.
Five questions that decide it
Forget the hype. Answer these five honestly and the decision usually makes itself.
- How often will a real customer open it? Weekly or more → app candidate. Monthly or less → website, no contest. Be ruthless here — founders overestimate this by 10x, every time.
- How will customers find you? If the answer is Google, you need a website — apps are invisible to search engines. If you already have an audience (email list, foot traffic, existing customers), an app has a fighting chance.
- Do you need the phone itself? Push notifications, camera, GPS, offline-first, biometrics, background location — that's app territory. Bookings, payments, forms, catalogs, content, chat — a modern browser does all of that.
- What's the year-two budget, not just the build quote? Add maintenance (15–20%), store fees, and the revenue cut. If year two makes you flinch, you can't afford the app — and that's fine, because most businesses can't.
- How fast do you need to learn? A website lets you change the offer on Tuesday and measure results on Wednesday. An app makes every change wait for store review — Apple rejects roughly 40% of first submissions and takes 24–72 hours per review cycle.
The verdict by business type
| Business | Build this | Why |
|---|---|---|
| Local services (plumber, dentist, salon, law firm) | Website | Customers Google you once every few years. Nobody downloads an app for a haircut. |
| Ecommerce store | Website first | Prove your repeat purchase rate first. An app earns its keep only when loyal customers order often enough to want the icon. |
| Restaurant / café | Website + online ordering | An app only makes sense for chains with a real loyalty program. |
| Creator / coach / consultant | Website | Your site is home base. Your audience lives on platforms you don't own — the site is the one asset that's yours. |
| Marketplace / two-sided product | Responsive web app | This is where most AI-built prototypes live. Validate the marketplace in the browser before you even think about native. |
| SaaS / software | Web app first | Go native when your analytics show mobile dominating usage — not before. |
| Fitness / habits / daily-use consumer | App can win | Daily habit + push reminders + offline use = the rare case where the app math works. |
The middle path nobody pitches you: the PWA
There's a third option, and agencies rarely mention it — because it's cheaper than what they'd rather sell you. It's called a progressive web app.
A PWA is a website that behaves like an app: installable to the home screen, works offline, sends push notifications. One codebase, no app store, no 30% cut. For a huge middle of businesses, it's 80% of the app experience at 20% of the cost.
It's not magic. iOS has historically limited what PWAs can do — push notifications and background powers have been the weak spots, though both are improving — and a PWA will never match native for heavy graphics or deep hardware integration. But if your "app idea" is accounts, content, bookings, messaging, or a dashboard, that's a PWA. Anyone quoting you native prices for that is either uninformed or upselling you.
The trap we keep rescuing people from
AI builders like Lovable and v0 made this trap ten times more common. You can now generate something that looks like a finished app in a weekend. Then reality arrives: it can't pass App Store review, the auth is held together with hope, the data model collapses at a hundred users, and there's no backend worth the name. We wrote a full engineering checklist on taking AI-built apps to production — the most honest thing we've published on the subject.
The pattern is always the same: build the flashiest thing first, validate last. Flip it. The browser is the cheapest validation lab ever invented. If the idea works as a web app, you've saved $50,000 proving it. If it doesn't, you've saved $50,000 learning it.
Sitting on an AI-built prototype right now? Our free rescue audit will tell you what's shippable, what's salvageable, and what's cheaper to rebuild — before you spend another dollar.
When you SHOULD build an app
To be fair — sometimes the app is the right call. Build one when:
- Customers will genuinely open it weekly or more — banking, fitness tracking, delivery, team tools, a school app parents check daily.
- Push notifications are the product — reminders, alerts, and messaging that lose all value sitting in an email inbox.
- You need offline-first or the hardware: field service, barcode scanning, camera capture, background location.
- You've outgrown the website: your analytics show 70%+ mobile traffic with heavy repeat visits, and users are asking for it. That's demand, not a hunch.
An app is a retention tool. If you have nothing to retain yet, it's a very expensive brochure.
The build order that actually works
- Ship the website. Four to eight weeks. Start collecting traffic, emails, and orders while the app-first crowd is still in sprint planning.
- Watch the data for three to six months. Mobile share, repeat visits, feature requests, support tickets — the data will tell you what to build next.
- If the data says "mobile-first audience," ship a PWA. Home-screen presence without the store tax.
- If the PWA's most-used features genuinely need native powers, build the native app — one platform first, funded by revenue the product already earns.
- Never start at step four.
Still torn? Here's the offer: book a call and describe what you're trying to build. We'll tell you straight whether you need an app, a website, or neither — including the times the honest answer costs us the project.
