CPI: Why A Falling Cost Per Install Can Be Bad News

The short version

CPI is what it cost to acquire one app install.

javascript

CPI = spend / installs

It is the entry-level efficiency metric in mobile user acquisition: easy to read, easy to optimize toward, and easy to be misled by, because it says nothing about whether the installs were worth buying.

The trap

CPI is a cost metric on an event that carries no revenue.

That single fact produces the most reliable way to lose money in mobile UA: optimizing CPI down while user quality falls. Your reporting improves while your business worsens, and nothing in the metric signals the difference.

Two campaigns can share a CPI for entirely different reasons. A creative with strong pull lowers CPI by raising install rate. A competitive auction raises CPI by raising CPM. Same number, opposite situations.

What CPI tells you

The immediate cost of volume.
Whether a channel is affordable at the top of the funnel.
How auction pressure is moving, when read alongside CPM.

What CPI does not tell you

Whether the user will open the app again.
Whether the user will ever pay.
Whether the install was incremental, or would have happened anyway.

A CPI is not affordable until step three

The relevant comparison is never CPI against an industry figure. It is CPI against what that cohort eventually returns.

CPI tells you what an install cost.
Retention tells you whether the install became a user.
ARPU and LTV tell you what that user is worth.
Payback period tells you when you get the money back.

Judging a CPI at step one is guessing. Setting a CPI target before you know what a user is worth is guessing with a number attached.

Reading the patterns

CPI rising, CPM rising. Auction pressure — competition, seasonality, or a narrowed audience. The creative may be fine.

CPI rising, CPM flat. Install rate is falling. Look at IPM and creative fatigue rather than at bids.

CPI falling, retention falling. The dangerous one. You are buying cheaper users who are worth less, and reporting looks like an improvement.

CPI stable, revenue per cohort falling. Audience quality is drifting even though acquisition cost is not.

Platform notes

On iOS, install attribution is constrained by the privacy framework, so install counts — and therefore CPI — can differ substantially between the ad platform and your MMP. Decide which source is authoritative before optimizing against the number.

On Android, attribution is generally more complete, which brings platform-reported and MMP-reported CPI closer, though still not identical.

Where to see it in Playad

Install volume by channel is available in Dashboards as a channel installs trend, and cost columns sit in Creative Performance. When an MMP is connected, the cohort views are what let you judge whether a given CPI was worth paying.

Common mistakes

Setting a CPI target before knowing what a user is worth.
Comparing CPI across geos, where both cost and user value differ.
Optimizing toward install rather than a downstream event, then being surprised by retention.
Treating a CPI improvement as a win without checking cohort quality.

Practical rule

Never evaluate CPI on its own. Pair it with retention and revenue for the same cohort, and treat a falling CPI as a question rather than an achievement until you have checked what you bought.