Show Me the Money

I · Cold open

A paywall that works. €0.

Bedtime Stories · August
Installed	49
Asked for a story	32
Read it to the end	25
Reached the paywall	15
Started a trial	0

Every event fired. Every screen rendered. Billing works. €0.00.

I · Cold open · beat 1 · 1:46 · 176 words

Hello, everybody. Today we look at how I tried to make money with my apps. So far, mixed results. I'm going to put a lot of numbers on this screen, and some of them are embarrassing. They are real numbers from my apps, pulled a few days ago.

In August, forty-nine people installed my bedtime-story app. I built it for my kid. It uses AI, and it puts your kid's name in the story, their friends, the places they like. Thirty-two of them asked for a story, twenty-five read it to the end, and fifteen reached the paywall. Not one started a trial.

And nothing was broken. Every event fired, every screen rendered, CI was green, billing works. Zero money.

I'm opening with my worst app because this is the honest state of most indie apps. On Twitter you see the indie dev who makes a fortune with a weird little app. Nobody shows this slide. And for a long time I didn't look at numbers like these at all. So first, how I started looking.

I · Who is talking

15 years of mobile.
Last year everything changed.

I · Who is talking

1 person. 14 apps. 0 team.

14apps on Android

13of them on iOS too

1.5Minstalls, one app

Every install organic, out of Play Store search. No budget, no marketing department, no cofounder. Releases scripted, metadata in the repo, social by a bot.

I · The claim

Money is a quality feature.

Downloads flatter you
Ratings flatter you
Revenue does not

To be paid you must measure. To measure is to find out.

II · The first look

57% of the money. 11% of the ads.

Format	Impressions	Revenue
Banner	48%	16%
Rewarded + rewarded int.	11%	57%

23,000 banner impressions → €27.
5,400 rewarded → €95.
Rewarded is the format I refused for a year.

II · The first look · beat 5 · 1:18 · 141 words

The first thing I did was sort my revenue by format. What makes money, and what is only noise. Ads are the simplest money you can add: drop in the SDK, add a banner, and money comes in.

In September my banners did twenty-three thousand impressions and earned twenty-seven euros. My rewarded ads did five thousand four hundred impressions and earned ninety-five euros. Eleven percent of the impressions, fifty-seven percent of the money.

I resisted rewarded ads for years, because I hate ads. I pay for YouTube, I pay for Duolingo, I pay so I don't see ads. But I'm not my user. My user has a hundred-euro phone, and if the reward is right, they choose to watch. A rewarded ad says: if you want something more, watch this and it's yours. That choice is why it pays a premium.

II · The first look

6 formats. 44× the price.

Format	eCPM
Native, badly placed	€0.54
Banner	€1.19
App open	€7.31
Interstitial	€8.24
Rewarded interstitial	€13.75
Rewarded	€23.64

Whole portfolio, September: same users, same month. A factor of 44.

II · The first look

Half the ads. The same money.

−58%ads per hour of play

+3%money per hour of play

Blackjack Android, 30 days either side of 18 August. The banner held 68% of impressions at €1.25 eCPM. People also played 46% longer; I cannot prove the banner is why.

II · The bugs were mine

Loading an ad is not showing an ad.

Kindle Gratis · 7 days	July	October
Rewarded ads asked for	1,461	560
Rewarded shown	18%	83%
Rewarded earned	€6.68	€12.14
Interstitials shown	19%	72%

Same price per rewarded ad: €25.51 then, €26.16 now. Weeks of 15–21 July and 29 Sep – 5 Oct.

II · The bugs were mine · beat 8 · 1:33 · 169 words

Then the bugs. This was an expensive lesson, and an AdMob consultant taught it to me. Every now and then Google gives you a call with one, and they spotted my problem straight away.

I was loading ads ahead of time. I thought I was clever: when the user gets to the ad, it's already there, no waiting. In one week in July my Kindle app asked for almost fifteen hundred rewarded ads, and showed eighteen percent of them. Every request starts an auction. Buyers bid on your slot, and then you don't show the ad. Do that enough times, and they stop bidding.

So I load an ad only when somebody is about to see one. In the first week of October the same app asked for five hundred and sixty, and showed eighty-three percent. It earned eighty-two percent more from rewarded, and the same number of people used the app in both weeks. I was sure I was doing the right thing. The numbers said I wasn't.

II · The bugs were mine

When the ad failed to load, the paid books were free.

if (rewardedAd.isReady) {
    showAd { unlockBook() }
} else {
    unlockBook()          // no ad? have it for free
}

Every failed load — bad network, no fill, expired — opened the entire premium catalogue. The leak only opens when monetization is already failing.

III · Switches, not releases

Every ad has a switch. None of them needs a release.

Remote Config key	Live	What it decides
ads_<format>_enabled × 4	true	an off switch for each format
ad_ri_frequency	500	hands between rewarded interstitials
ad_interstitial_frequency	20	hands between interstitials
ad_app_open_daily_cap	1	app open ads per day
ad_app_open_min_hands_played	20	hands before the first app open ad

Blackjack Trainer, live values. Every value is clamped in the app; the switches mirror defaults in the code, so a Remote Config outage leaves the ads on. The rewarded interstitial made 81% of the app's ad revenue in September.

III · Switches, not releases

I switched 1 ad off. Every interstitial stopped.

4 October, CEST	Remote Config
21:56	rewarded interstitial: every 10 hands → every 20
22:02	rewarded interstitial: off
22:02	…and every interstitial stops, on each phone as it fetches
22:40	workaround: on again, every 500 hands
next day	the real fix, merged in the code

A spacing rule read a disabled format as one that was just shown. The off position of a switch is a code path. Test it.

III · Switches, not releases · beat 11 · 1:34 · 169 words

But things break. On the fourth of October I was playing my own blackjack app, and I got a full-screen ad roughly every two minutes. So I did what the switches are for. Rewarded interstitial, from every ten hands to every twenty. Six minutes later I switched it off completely. I lose some money, but I stop annoying the players. And in September that format made eighty-one percent of the app's ad money.

And every interstitial in the app stopped, even the ones I didn't touch. A spacing rule, never two full-screen ads too close together, read the format I'd switched off as one that had just been shown.

Thirty-eight minutes later the workaround was live: switch it back on, once every five hundred hands. That held for the evening, and the machine wrote the real fix, merged the next day. The off position of a switch is a code path. Test it.

That's ads. In September they made three quarters of my money. The rest came from subscriptions.

IV · Subscriptions

€2.99 a week. 0 sold so far.

Bedtime Stories	Price	Trial
Weekly — listed first	€2.99	3 days
Yearly	€34.99	7 days
Monthly	€4.99	7 days

Weekly available since 26 September, listed first in the app. One Italian price, equalised across 175 App Store territories — for now.

IV · Subscriptions

2 apps. 2 paywalls. 1 sells.

	3 Things a Day	Bedtime Stories
Active subscriptions	7	0
MRR	€22	€0
Who sees the paywall	everyone who finishes onboarding	15 of 25 readers
Tap to buy	13% of views	0 trials

IV · Subscriptions · beat 13 · 1:24 · 153 words

Both apps use RevenueCat. Three Things a Day has seven active subscriptions and twenty-two euros a month. The bedtime app has none.

Three Things a Day shows its paywall at the end of onboarding, to everyone who finishes it. On Android, in August and September, thirteen percent of the paywall views led to a tap to buy. In the bedtime app, in August, fifteen of the twenty-five parents who finished a story saw the paywall, and none of them started a trial.

They are different apps. Three Things a Day is a habit app. You install it because the idea sold you, you finish onboarding, you see a trial, and you start it before you have even used the app. Bedtime Stories is entertainment. You read a story, your kid falls asleep, and you go to the other room. I can't prove which difference matters. But a paywall that nobody sees sells nothing.

IV · Experiments

We are in the middle of it.

Bedtime: does weekly-first sell? First numbers by November
Bedtime: the paywall at the end of the first story
Prices by country, not one Italian price everywhere
Smart Pantry Plus, no ads: 2 yearly in its first week

One answer so far: 4 Plus sold, €74. The app's ads made €5 in all of September.

IV · Experiments · beat 14 · 1:25 · 154 words

So we are in the middle of it. Four experiments, running or about to start.

The weekly plan: no sales yet, and I'll know more in November. The paywall at the end of the first story, when your kid is asleep and you have just seen the app work. Prices by country, because one Italian price is maybe too high in some places. And a subscription for my pantry app. People complained about ads everywhere, one of them in a one-star review, and the app runs AI, so it costs me money to run. Smart Pantry Plus removes the ads and adds more on top. It went live on the App Store on the eleventh of October. That week, two people bought the yearly plan, for forty-five euros. Then the first one on Android, a monthly, and on Tuesday a third yearly. The ads in that app made five euros in all of September.

IV · Experiments

10 minutes beat a whole day.

Kindle Gratis · 1 ad opens VIP for	10 minutes	A whole day
Rewarded ads per user	1.83	0.74
Ad money per user	8.6¢	3.2¢
Back the next day	34%	32%

A/B test, 2–8 October. Same retention, 63% less money. Stopped.

V · Who checks my work

No QA team. A review panel that never gets tired.

A plan — reviewed by other models before any code
My approval
A pull request — reviewed by models that did not write it
Fix every blocker, review again
I merge. Nobody else.

ship-feature

1,750 pull requests, 50+ repositories, since June.

V · Who checks my work · beat 16 · 1:22 · 150 words

I'm alone, but it's 2026, so I'm not really alone. I built two tools for it: ship-feature and pr-review-relay.

It starts with a chat with Claude Code or Codex, and we end up with a plan. ship-feature sends the plan to Codex, Grok and GLM, and they argue with Opus until they agree. Then it comes to me. I approve it, and Opus writes the code on a branch, as a pull request.

Then pr-review-relay asks the same three models to review the pull request. They go back and forth with Opus until there are no blockers and nothing left that should be fixed. Then they ask me to merge, and I merge. Only me.

Since June, seventeen hundred and fifty pull requests went through it, in more than fifty repositories. I get better code in less time, and I get time back for new apps and for my family.

V · Who checks my work

1 review.
2 broken promises.

Two Notes	The screen said	The code would
Rewarded ad	Watch it, get a life	End the run anyway
Paywall	Free trial	Charge for a year

Both found by Codex in one review round, before launch.

V · Who checks my work

The build refuses to ship.

Built from the tip of the release branch
Unit tests green
Maestro on a real phone, through the path that pays
Release build: no “Test Ad” on screen, ever

Red gate, no upload. My fake ad was once more forgiving than the real one — the test passed while the bug was live.

V · The machine that watches

1 text file. 30% of the revenue.

€5.00eCPM, before

€3.50eCPM, after

0warnings

app-ads.txt broken in a site rebuild. Traffic flat, app unchanged. Programmatic buyers verify it, then quietly stop bidding. I saw it only because I happened to look.

V · The machine that watches

I stopped opening dashboards. They come to me.

When	Job	Speaks when
07:00	Revenue, users, subscriptions, reviews	every day — one message
08:00	Crashlytics, every app	only past a threshold
every 15 min	Sentry, every backend	only if new, regressed or escalating
Monday	Dependency updates	a PR per repo, tests green
any time	A job that dies	its name and its log

app-tools: one CLI over AdMob, GA4, Search Console, both stores and RevenueCat — and an MCP server.

V · The machine that watches

The month my best format stopped firing.

KG · app open, 30 days	Before 15 Aug	After
Revenue	€47	€10
Impressions	4,694	1,036
Shown	40%	11%
eCPM	€9.98	€9.68

The price never moved. The ad stopped being shown. −79%, about €37 a month, for six weeks.

V · The machine that watches

7 in the morning. 1 message.

Show rate — filled → shown

Smart Pantry (A)    app open        503 →   8     2%
Ebooks for Kindle   app open       5726 → 606    11%
StreakUp (iOS)      interstitial    395 → 105    27% ⚠

⚠ = interstitial under 30% with 300+ requests:
    the app loads them and never shows them

From the daily report of 21 September, translated from Italian. The warning covers interstitials only — which is why the Kindle line never got one. Same week: a backend job that failed ~680 times and reported success became the Sentry job.

VI · The doors

An ad unit with no users pays exactly €0.

€24eCPM

× 0users

= €0revenue

Everything before this multiplies one number: the people inside the app. Two doors I can work on: the store listing, and what the apps post. I have never paid for an install.

VI · The doors

Conversion 26% → 36%, same app.

The icon
The first two screenshots
The first line of the description — the only line anybody reads

26.3% → 35.7% install conversion. A third more users, every day, forever, for one afternoon. Not a feature — a better sentence.

VI · The content machine

Something goes out every day. A machine makes it.

Who	What	Where
Kindle Gratis	carousel: today's 3 free books, from the app's data	TikTok · EN + DE
3 Things a Day	promo reel, one caption per network	TikTok · X · LinkedIn
One app a day	daily standup, from git	X · LinkedIn
Me	what I earned last month: €215 in September	X · LinkedIn

Rendered with HyperFrames, published with Buffer. Reels: github.com/hamen/app-promo-reel (MIT).

VI · The content machine

The hook moved to frame 0.

~1 saverage watch time

29 → 240views, video A

111 → 244views, video B

Same two videos, old hook vs new. A tiny sample — posted on different days. Still testing.

VII · What it made me fix

What the money made me fix

What money made me fix
Login wall + 4-screen onboarding	5 of 206 reached the story
Startup time	a slow first launch gets no second
Paywall copy	13 views, 0 taps
Analytics	2 years of guessing

Not one of these is monetization work. Every one came out of looking at money.

VII · What it made me fix

You cannot be paid by a user who wants nothing.

57%of the money

11%of the impressions

Rewarded + rewarded interstitial. A rewarded ad is a trade — nobody can be tricked into it. My best format is a measurement of whether I built something worth wanting.

VII · What it made me fix

Money finds the bug. It will not fix it.

900+customers, lifetime

€0revenue, lifetime

3paywall rewrites

Revenue is the best defect detector I have ever had, and a terrible engineer. It finds the bug, then stands there looking at you.

VII · Close

The checklist

Check app-ads.txt today — the cheapest alarm you will ever install
Make the numbers come to you — one message, every morning
Sort revenue by format, not by impression count
Put every ad behind a switch — and test the off position
Let a model that did not write the code review it
Show the paywall to everyone who could buy
Count every step to the paywall, and find where people leave

Bedtime Stories, April: 5 of 206 reached the story.

VII · Close

Ask me the uncomfortable ones.

ivanmorgillo.com
indiepa.ge/ivanmorgillo — every app, numbers included

Every number here came off my own dashboard, including the ones that embarrassed me.

Why I looked

A paywall that works. €0.

15 years of mobile.
Last year everything changed.

1 person. 14 apps. 0 team.

Money is a quality feature.

What the money said

57% of the money. 11% of the ads.

6 formats. 44× the price.

Half the ads. The same money.

Loading an ad is not showing an ad.

When the ad failed to load, the paid books were free.

The controls

Every ad has a switch. None of them needs a release.

I switched 1 ad off. Every interstitial stopped.

The paywall

€2.99 a week. 0 sold so far.

2 apps. 2 paywalls. 1 sells.

We are in the middle of it.

10 minutes beat a whole day.

How one person ships

No QA team. A review panel that never gets tired.

1 review.
2 broken promises.

The build refuses to ship.

1 text file. 30% of the revenue.

I stopped opening dashboards. They come to me.

The month my best format stopped firing.

7 in the morning. 1 message.

Growth

An ad unit with no users pays exactly €0.

Conversion 26% → 36%, same app.

Something goes out every day. A machine makes it.

The hook moved to frame 0.

Back to the claim

What the money made me fix

You cannot be paid by a user who wants nothing.

Money finds the bug. It will not fix it.

The checklist

Ask me the uncomfortable ones.

Grazie.

Show Me the Money

Why I looked

A paywall that works. €0.

15 years of mobile.Last year everything changed.

1 person. 14 apps. 0 team.

Money is a quality feature.

What the money said

57% of the money. 11% of the ads.

6 formats. 44× the price.

Half the ads. The same money.

Loading an ad is not showing an ad.

When the ad failed to load, the paid books were free.

The controls

Every ad has a switch. None of them needs a release.

I switched 1 ad off. Every interstitial stopped.

The paywall

€2.99 a week. 0 sold so far.

2 apps. 2 paywalls. 1 sells.

We are in the middle of it.

10 minutes beat a whole day.

How one person ships

No QA team. A review panel that never gets tired.

1 review.2 broken promises.

The build refuses to ship.

1 text file. 30% of the revenue.

I stopped opening dashboards. They come to me.

The month my best format stopped firing.

7 in the morning. 1 message.

Growth

An ad unit with no users pays exactly €0.

Conversion 26% → 36%, same app.

Something goes out every day. A machine makes it.

The hook moved to frame 0.

Back to the claim

What the money made me fix

You cannot be paid by a user who wants nothing.

Money finds the bug. It will not fix it.

The checklist

Ask me the uncomfortable ones.

Grazie.

15 years of mobile.
Last year everything changed.

1 review.
2 broken promises.