Cost per install (CPI)
Acquisition metric. How do users find and install the app?
What is cost per install?
Cost per install (CPI) is the average amount an app pays in advertising for each install a campaign drives: ad spend divided by the installs attributed to that spend.
Attributed install: an install credited to a specific campaign, ad or link, as opposed to an organic install that no tracked source is credited with.
Self-attributing network (SAN): an ad network such as Meta or Google that decides for itself which installs it drove and reports them to the MMP, which then deduplicates them.
How to calculate CPI
CPI = Ad spend ÷ Attributed installs- Ad spend
- media cost for the campaign, channel or period, excluding tax if your spend reports exclude it
- Attributed installs
- installs credited to that spend by the ad network or a mobile measurement partner (MMP)
Where definitions differ
| Variant | Formula | When to use it |
|---|---|---|
| Paid CPI | Paid spend ÷ Paid-attributed installs | The usual meaning in campaign reports, and the one to use for channel decisions. |
| Blended CPI (often called eCPI) | Total spend ÷ All installs, organic included | Always lower than paid CPI when an app has organic installs. Useful for budget planning, misleading for comparing channels. |
| CPI as a price | The fixed amount agreed per install on a CPI-priced network | Here CPI is what you pay per install by contract, not an average worked out afterwards. |
By channel: Splitting CPI by channel, campaign or creative needs installs that are attributed and deduplicated across ad networks, which is the job of MMPs such as Linkrunner, AppsFlyer and Adjust.
CPI calculator
Example inputs: change them to your own numbers. The examples only make the calculator show a worked answer; they are not benchmarks or targets. Runs in your browser; nothing you type is sent anywhere.
- CPI
- $2
- Highest CPI that still meets your target cost per sign-up
- $2.50
What is a good target cost per install?
A target CPI should come from what an install is worth to you, not from a market average: multiply your target cost per sign-up (or per payer) by the share of installs that reach that step. The second line of the calculator does exactly this sum.
How CPI differs by vertical
| EdTech | Read next to cost per sign-up or trial, because many learners install and never start a first lesson. |
|---|---|
| Fintech | Weak on its own. The cost that matters is per KYC-complete user or per first transaction. |
| OTT and streaming | Free tiers make installs cheap; compare CPI with cost per subscriber before scaling a channel. |
| Gaming | The headline buying metric, but always read against D1 retention and D7 ROAS, because cheap installs from poor inventory rarely pay back. |
For published CPI medians by app category in India, see the CPI benchmarks on AttributionBenchmarks.
Common mistakes
- Dividing spend by all installs, organic included, and reporting it as paid CPI.
- Comparing CPI across ad networks that count installs on their own rules and windows without deduplicating them first.
- Mixing currencies, or tax-inclusive and tax-exclusive spend, in one figure.
- Pushing CPI down while activation falls, so the cost per real user goes up.
Questions
How do you calculate cost per install?
Divide the ad spend for a campaign or period by the installs attributed to it. For example, $20k spent for 10k attributed installs is a CPI of $2.
What is the difference between CPI and eCPI?
CPI usually means paid spend divided by paid installs. eCPI is the effective cost per install worked out after the fact, often across all installs including organic ones, so it comes out lower.
Why does my ad network show a different CPI from my MMP?
Each counts installs with its own rules. Ad networks can include view-through and modelled conversions and use their own attribution windows, while an MMP deduplicates across networks, so the same spend is divided by different install counts.
Are the calculator's example inputs benchmarks?
No. The example inputs only make the calculator show a worked answer when the page loads; they are not typical, good or target values. Replace them with your own numbers.
Does the calculator send my numbers anywhere?
No. It runs entirely in your browser, and nothing you type is stored or sent anywhere.
Why do tools report different values for the same metric?
Because definitions differ: what counts as active, which attribution window applies, whether organic users are included and how days are cut by timezone. Align the definition before comparing numbers.