What is a good cost per install?

Any CPI below what a user is worth to you over their lifetime. That makes it an entirely relative figure: eight dollars is excellent for a fintech app and ruinous for a casual game. Compare against your own LTV, not against a published average.

Why is iOS CPI so much higher than Android?

Higher purchasing power and higher competition for the same inventory, so the auction clears higher. Reported averages currently put iOS around three times Android, which broadly matches the difference in revenue per user.

Is my reported CPI accurate?

It is accurate about what the platform measured. It does not account for taps that never reached the store, which on iOS social traffic is a real share. Compute the version using your store console numbers and compare.

How do I lower my cost per install?

In order of return per hour spent: fix delivery so more taps arrive, improve the first two screenshots so more store views convert, then iterate creative. Bidding changes are the last lever, not the first.

Cost per install explained

CPI is the most quoted number in app marketing and one of the least reliable, because the denominator counts things that may never have happened.

9 August 2026 · PAD team

The definition, and the version that is true

Nominally, cost per install is spend divided by installs. In practice the figure most teams look at is spend divided by installs the platform attributes to itself, which is a different number and generally a friendlier one.

reported CPI   = spend / installs the platform claims
effective CPI  = spend / installs your store console actually shows

The gap between them comes from attribution windows, from cross-device claims, and from taps that never reached the store at all. Only the last of those is something you can fix yourself.

What CPI looks like now

Reported industry averages put iOS around six dollars and Android around two, with enormous variance by category. Casual games sit at the low end, finance and subscription apps several times higher, and tier one markets run three to five times emerging ones. Published sets like Business of Apps are fine for orientation and misleading if used as a target.

Category	Rough iOS range	Why
Casual games	low	broad audience, high tolerance
Utilities	low to mid	clear promise, quick decision
Subscription health and fitness	mid to high	narrow audience, high intent required
Finance and fintech	high	small qualified pool, high LTV justifies it

Your own number from last month is worth more than any of these. Benchmarks tell you whether you are in a plausible range, not whether you are doing well.

The silent inflation

If a share of your taps never reaches the store, you still paid for them. That does not raise your reported CPI, because the platform did not count those as installs either. It raises your effective cost of a working install, quietly, by the size of the leak.

Some arithmetic. Suppose you spend a thousand dollars and the platform reports two hundred installs at five dollars each. If 15% of taps never arrived, you paid a thousand dollars for a path that could only ever deliver 85% of what you bought. The five dollar figure is real. It is just not the whole price.

How to compute the honest version

Take spend for a clean period, one channel, one country.
Take installs from the store console rather than the ad platform, for the same period and geography.
Divide. That is your effective CPI.
Compare against the reported figure. A persistent gap is a measurement problem worth an afternoon.

What to do with the number once you have it

CPI on its own is not actionable, because a cheap install that churns in a day costs more than an expensive one that stays. Pair it with day seven retention and you get cost per retained user, which is the figure worth optimising and the one almost nobody reports.

The denominator is easier to fix than the numerator.

Create a link