Type at least 3 letters to search

·18 min read

Mobile App Marketing: What It Costs to Get People to Use Your App

App marketing cover: a handmade violin and bow lying in a spotlight on an empty theatre stage, facing rows of empty red velvet seats

You really can put an app together with AI over a few evenings today. Getting people to find it, install it and open it a second time takes as much money as it always did, and often more. Here is why an app still costs money, how much to set aside for marketing, what the market figures show, and how to test yourself before you spend six months.

SHORT ANSWER

The short answer

Budget at least as much for first-year marketing as the app itself costs at market price. Not what you spent on it, but what it would cost if a team built it. In the US market that is roughly $100,000 for the app and the same again for promotion, and that is not even a high figure. If you vibe-coded your app for free, the marketing does not become free with it.

The price of an app sets whether you can even try to work in this market. The build buys you a ticket in, and the marketing budget decides whether anyone sees you once you are inside. If you have money for neither the app nor its promotion, think a thousand times before you start: a website, an email list or a messaging channel may do the job.

The "at least 1:1" rule comes from our own practice; there is no hard research on it. The figures below suggest it is on the cautious side: $100,000 of US advertising earns back roughly a fifth to a third of the spend in the first year, counting the revenue from the people who start paying.

WHY IT COSTS

Why an app still costs money when AI writes the code

Models such as Claude, Codex and Cursor get noticeably better every month. What took weeks two years ago, an experienced developer working with AI now does in days. So it feels as if an app should cost nothing. In fact, AI made typing code cheap, and most of what you pay for is something else.

  • Verification. Someone has to understand what the model wrote and answer for it. We wrote separately that the cost of AI is in verification, not generation.
  • Architecture. It decides whether the app survives its hundredth feature and its thousandth user. Code made of patches stops taking changes exactly when growth arrives, as we explain in the hidden cost of the wrong architecture.
  • Security and data. Keys left in the code and an open database full of customer data. The owner answers for a leak, not the model.
  • Server and payments. What the screen does not show: the load on a Friday night, or a payment that hangs just as the customer is ready to pay.
  • Stores and rules. Apple and Google review, a privacy policy, account deletion, updates for every new iOS and Android version.
  • Time. Six months of evenings is a cost too; nobody just counts it.

That is why a designer who does not understand architecture ends up with an app whose market value tends toward zero: it cannot be developed safely and has to be rewritten from the start. How to decide what to do with a version like that is covered in our article on an MVP built with AI. What an app built by a team costs, and what goes into that price, is in how much mobile app development costs.

The conclusion for a vibe coder is simple: code got cheaper, while architecture, responsibility and marketing did not. Marketing is now becoming the biggest line in the budget.

AN EXAMPLE

A typical mistake: payments on a Friday night

Here is an example of how a vibe-coded app differs from one written properly. Take food delivery and its most important moment, the payment. In tests with ten orders both versions behave the same. The difference shows up on a Friday night, when orders come in one after another.

Typical code that AI happily writes when asked to "make the order paid after payment" looks roughly like this. The code is simplified, and it is in Laravel, which we most often use for app servers.

// What vibe-coded code often looks like: the phone itself tells the server the order is paid
Route::post('/orders/{order}/paid', function (Order $order) {
    $order->update(['status' => 'paid']);
    Mail::to($order->email)->send(new OrderPaid($order)); // a slow call right inside the request
    Http::post($kitchenUrl, $order->toArray());           // and another one
    return ['ok' => true];
});

There are three problems here. First, anyone who finds this address can mark any order as paid without paying, because the server takes the phone at its word. Second, the customer email and the kitchen message are sent during the request itself, so each request waits several seconds. On a Friday night the server quickly runs out of worker processes, the app shows an endless spinner, and the customer taps "pay" again. Third, nobody checks the repeated request, so a second payment and a second kitchen order are possible.

A properly written version is built differently.

// How it should be: the payment provider tells the server, with a signature
Route::post('/webhooks/payments', function (Request $request) {
    $event = PaymentWebhook::verify($request);  // a request without a valid signature is rejected
    $new = ProcessedEvent::firstOrCreate(['event_id' => $event->id])->wasRecentlyCreated;
    if ($new) {
        FulfillOrder::dispatch($event->orderId);  // kitchen, email and push go to a queue
    }
    return response()->noContent();  // answers in milliseconds, repeats are ignored
});

The payment provider itself tells the server about the payment, and the server checks its signature, so a payment cannot be faked. Each event is recorded by its ID with a unique key in the database, so a repeat delivery of the same event breaks nothing. Everything slow, the kitchen, the email and the notification, goes to a job queue and runs separately, while the payment provider gets an answer in milliseconds. This is exactly what Stripe recommends in its webhook documentation: verify the signature, return a success response quickly before any complex logic, and track event IDs to spot duplicate deliveries. For requests to the payment provider itself there are idempotency keys, which let you safely retry a request without performing the same operation twice. The app on the phone, meanwhile, never decides whether an order is paid: it only asks the server for the status.

The difference in code looks small, and in money it is huge. The first version works until the first busy evening or the first person who looks at the app's requests. And that happens exactly when you have already paid for ads to bring those people in.

THE FLOOD

Twice as many apps, the same number of people

Sensor Tower estimates that about 560,000 new apps were released on the App Store in the first half of 2026, twice as many as a year earlier. Downloads over the same period grew by only 2% (as reported by 9to5Mac). In the third quarter, global downloads grew by 0.1% (Sensor Tower). New subscription apps, according to RevenueCat, launched at about 2,000 a month in early 2022 and at more than 14,700 a month by January 2026.

Analysts link this flood to vibe coding tools, although nobody can prove it directly yet. For a founder the consequence is the same: the number of people downloading anything has barely moved, while the number of competitors for their attention has doubled.

Apple has already responded. In June 2026 it rewrote guideline 4.3(b) of its App Store Review Guidelines: do not submit apps indistinguishable from what is already widely available, and Apple may remove such apps if they are not updated, improved or do not attract customers. In other words, a clone without customers can simply disappear from the App Store.

Code got cheaper. People's attention did not.

1:1

The "at least 1:1" rule and what $100,000 buys

The formula is simple: budget at least the market price of your app for first-year marketing. Not what it cost you. We count the market price the way a team building the app for a client would, with design, both platforms, a server and store publishing. In the US that is easily $100,000, so $100,000 is also the marketing figure to hold.

Why at least that much? Look at what the money buys. The calculation is ours, based on published market data, and it is approximate: prices and conversion depend heavily on category, country and product quality.

US cost per installInstalls for $100,000Started paying (2.6%)Cost per payerTheir first-year revenue ($32 per payer)
$2.51about 39,800about 1,040about $97about $33,000
$3.28about 30,500about 790about $126about $25,000
$5.0020,000520about $192about $17,000

Where the numbers come from. $2.51 is the average cost per download in Apple Ads in the US, from Adapty data as of July 2026. $3.28 is the cost of a shopping app install in the US in the first half of 2026, from Adjust. $5 is the top of the range for fitness apps in high-income markets, which Singular puts at $2 to $5. The share of people who pay and the first-year revenue per payer come from RevenueCat's report for North America: a median of 2.6% by day 35 and $32 for the year.

What follows from it. $100,000 of advertising brings 500 to 1,000 people who paid something and earns back 17% to 33% of the spend in the first year. That is optimistic, since the RevenueCat median includes people who came on their own without ads, and paid audiences usually do worse. It pays off only when people stay for a second and third year, which means the product has to be one people want to come back to.

So 1:1 is not "a lot for marketing." It is the amount at which things only begin. For scale: according to AppsFlyer, companies spent $109 billion on app marketing in 2025, $78 billion of it on user acquisition, with 42% of that money spent in the US.

ADS

Why ads at a few dollars a day do not work

The most common illusion: let's run ads at five dollars a day and see. In practice, ad platform algorithms do not even start working at those amounts.

Meta says it plainly in its help article on the learning phase: an ad set exits learning once it can deliver stably, which usually happens after about 50 results in a week. With fewer results, the ad set gets stuck as "Learning limited." In the same article Meta advises setting a budget large enough to get the results needed.

Let's do the math. If the result is an install at $3, then 50 installs a week cost about $150 per ad set. If you want the algorithm to find people who pay, and one payer costs about $150, then 50 payers a week come to about $7,500 a week. Google's best practices for App campaigns recommend a daily budget of at least 50 times the bid for install campaigns and at least 10 times for in-app action campaigns. At a $3.28 bid that is about $164 a day, or roughly $60,000 a year for a single campaign.

We wrote more about budgets in what Meta Ads budget you really need and why the learning phase deserves time to finish. In short: ads at a few dollars a day give the algorithm no data, and the money simply burns.

WHAT SELLS

What actually brings people into an app

We have built apps for delivery, for a large online store and for well-known restaurants. We cannot name them, but they share the same lessons. They were promoted with ads painted on delivery cars, targeted ads on Instagram and Facebook, and search ads. There were no ads inside the App Store or Google Play. Simple marketing moves sometimes helped, such as "install the app and get 25% off your first order."

But the product worked best. When the food is tasty and well made, people want to order again, and that is when they install the app. One detail matters: people mostly installed the app not on their first order but on their third or fourth. First a person trusts your kitchen or your store, and only then gives your icon a place on their phone.

Two conclusions follow. First, an app does not create demand. It serves the people who already come back to you more conveniently, so its cheapest channel is your own base: people who have already ordered, your website and your couriers. Second, the app store works as a search engine too. Apple says that almost 65% of downloads happen directly after a search on the App Store. The name, icon, screenshots and description on your store page matter, and they are worth testing: Apple lets you test up to three alternative product pages, and Google Play up to two variants.

Sometimes the marketing is run by a different team from the one building the app. We have worked both ways: sometimes we ran the promotion, and sometimes it was an agency the client had worked with for years while we handled development. What matters is that the two teams talk to each other before launch, not after.

REVIEWS

Reviews, ratings and crashes

People read reviews, especially for food. Someone ordering for the first time wants to know that others tried it and did not regret it. But ratings matter for more than people.

Apple says that ratings and reviews appear on your product page and in search results and can influence how your app ranks in App Store search. You can ask for a rating through the standard prompt no more than three times a year, so choose the moment: after a successful order, not on the first screen. You can answer every review, and the reviewer is notified and can update their rating. Google Play shows a rating weighted toward recent ratings, so old five-star reviews will not save you if the latest update broke everything.

For a vibe-coded app there is a hidden risk here. Google Play watches stability: if the share of sessions in which the app crashes exceeds 1.09%, or it freezes more often than the threshold, Play may reduce the visibility of the app and show a warning on its listing. An app that crashes on half of older Android phones loses its place in the store before you have paid for a single ad.

AFTER INSTALL

What happens after the install

An install is only the beginning. According to AppsFlyer data for the third quarter of 2026, very few of the people who install an app still open it a month later.

CategoryDay 1 (iOS / Android)Day 7Day 30
Food and drink15.5% / 13.5%6.8% / 6.0%2.5% / 2.3%
Shopping21.9% / 17.1%9.4% / 7.7%3.5% / 3.1%
Health and fitness21.7% / 17.0%9.7% / 7.4%3.6% / 3.1%

Out of every hundred installs, two or three remain after a month, four at best. One more AppsFlyer figure: in 2024, 46.1% of Android apps were uninstalled within 30 days of install, most of them on the first day.

Notifications help bring people back, but they have limits too. On Android 13 and newer, an app's notifications are off by default and permission has to be requested, and Google advises asking at a moment that makes sense, for example when the user places a food delivery order. According to Airship, about half agree: a median of 53% on Android and 49% on iOS. People who agree are quick to turn notifications off when there are too many.

So app marketing is not only paying for installs. It is also the work with people who have already installed: the first open, the first order, a timely and relevant notification, a reason to come back next week.

FITNESS

A story we hear every month

Most requests like this come from fitness. "We have 70 clients. We'll launch an app with video lessons and it will take off." The app is already vibe-coded or nearly done, and nothing stands behind the plan: no strategy, no ad budget, no idea where new people will come from.

When we explain the situation honestly, people often take offense, which is understandable: everyone wants to hear that the idea will work. They launch on their own, and a month or two later they see what we told them: without strategy, budget and marketing, an app does not grow by itself.

The market figures are unforgiving here. According to RevenueCat data for fitness apps, a median of 2.9% of people who download an app go on to pay. For 70 clients who already know you the share will be higher, but 70 people are a base for testing an idea, not a market. An honest first step for that studio is to give those 70 people the video lessons in the simplest way, through a members area on the website or a channel, and see how many watch every week and pay. If the results are there, then it is time to price the app and the marketing by the 1:1 rule.

SECURITY

Before you pay for ads, check security

Ads bring people, and people bring their data. If the app was built without understanding security, advertising only speeds up the moment something leaks.

There are examples already. In spring 2025, researcher Matt Palmer checked 1,645 projects built on the vibe coding platform Lovable and found databases reachable from outside in 170 of them: names, phone numbers, keys, payment details. The vulnerability was scored 9.3 out of 10, although Lovable disputes the rating. In February 2026, Wiz found a database key right in the page code of the vibe-coded service Moltbook: 1.5 million tokens and 35,000 email addresses were exposed, and the researchers call it a recurring pattern in such applications. And in Veracode's 2025 study, AI models introduced vulnerabilities in 45% of test cases, keeping in mind that Veracode itself sells security tools.

So before you put money into ads, show the app to people who understand architecture and security. That is what our MVP review from $3,000 is for: what in the code can stay, where the holes are, and what happens when there are a hundred times more users. It costs less than leaking the data of your first thousands of customers, the ones you already paid ads to bring in.

SELF-CHECK

How to test yourself before you start

Before you spend six months and real money, give yourself honest answers to a few questions.

  1. Do you have at least the app's market price for first-year marketing? If not, think a thousand times before you start.
  2. Why will someone come back every week? Without an answer, the app gets deleted in the first month, like half of all others.
  3. Who are your first hundred users and where are they now? Your own base and customers who already pay are the cheapest start.
  4. Who will answer reviews and watch the crash rate? Ratings and stability affect visibility in the stores.
  5. Has anyone checked the security of the code? Before the ads, not after the leak.
  6. Would a website do the job? If people come to you rarely, a fast website and search will do more than an app. See our article on when a business actually needs a mobile app.

If most of these questions have no answer yet, that is not a verdict on the idea but a hint about where to start. In a strategic session ($500, credited toward the project within 30 days) we work out the app, the marketing and the first-year plan together and tell you honestly whether to start now. And if launch is close, our Meta Ads management team can plan the budget so the algorithm gets enough data.

FAQ

Frequently asked questions

How much does it cost to market a mobile app?

We advise budgeting at least the market price of the app itself for first-year marketing. In the US market that is roughly $100,000. By our calculation on market data, that amount buys 20,000 to 40,000 installs and 500 to 1,000 paying users.

If I built my app with AI for free, can I spend less on marketing?

No. The 1:1 rule is counted from the app's market price, not from what you spent. The cost per install and the cost per paying user are the same for everyone who buys ads, however the code was written.

Why do ads at a few dollars a day not work?

Algorithms need data. Meta says an ad set exits the learning phase after about 50 results in a week, and Google recommends a daily budget of at least 50 times the bid for install campaigns. At a few dollars a day, ads never even start learning.

How many people are still using an app after a month?

Very few. According to AppsFlyer data for the third quarter of 2026, 2.3% to 3.6% of people who installed an app still open it after 30 days, depending on category and platform. Half of Android apps are uninstalled within the first month.

Do reviews affect downloads?

Yes. Apple says ratings and reviews can influence how an app ranks in App Store search, and Google Play weights its rating toward recent reviews and may reduce the visibility of an app that crashes often. Answer reviews and ask for a rating after a successful action, not at launch.

What if there is no money for marketing?

Start without an app. Test the idea through a website, a members area or a channel with people who already know you, and count how many come back and pay. Price the app and the marketing once there is proof the product is wanted.

Sources

Planning an app, but not sure there is enough money for people to hear about it? Let's work out the numbers together.

Book a Strategic Session

Meta Ads Management