Definition
Cost per install divides advertising spend by the installs credited to that spend. An install, for this purpose, is usually the first open of the app after it was downloaded, attributed to an ad by the network that served it or by a mobile measurement partner that applies one set of rules across networks. A download that is never opened does not appear, and an install that cannot be matched to an ad is counted as organic rather than paid.
Spend ÷ Installs(currency)
- Spend is the media cost of the campaign, ad set or network in the period. If agency fees or creative production are included, say so, because platform-reported CPI excludes them.
- Installs are the attributed first opens in the same period, under one attribution source. The network's own count and the measurement partner's count will rarely agree; pick one per report.
CPI is a narrow form of CPACost of each conversion. in which the action is the install. The difference matters because the install is not the action the business needs: it needs the person to register, complete a first task, subscribe or buy.
Why the count is less exact than it looks
On iOS, privacy rules mean installs from ads arrive through SKAdNetwork in delayed and aggregated form, while on Android a referrer passes campaign data into the first open. Modelled conversions fill some of the gap. The upshot is that CPI on one platform and CPI on another are built from different evidence, and the app attribution guide explains what each source can and cannot see.
How to interpret it
Lower is better only when the installs are of the same quality. The useful reading carries CPI forward:
- Cost per activated user. Divide the same spend by the installs that reached the app's first meaningful action, such as completing setup or a first use. This is the number that tells you whether cheap installs are worth having.
- Cost per payer. For apps that sell subscriptions or in-app purchases, divide by the installs that went on to pay. This moves CPI into the territory of CACCost to acquire one new customer..
- By network and campaign type. A network that sells installs cheaply to people who never open the app twice can be dearer per useful user than one with a higher CPI.
The app install campaigns guide explains why optimising to installs and optimising to an in-app event find different people, and why the second usually raises CPI while lowering the cost of a user who stays. The CPA calculator tutorial for app installs does the conversion from cost per install to cost per active user with your own figures.
A birding app follows installs into the first week
Brindlemoor Birding App spends £3,000 on each of two ad networks in a month. Network A delivers 3,000 installs, a CPI of £1.00; Network B delivers 1,500 installs, a CPI of £2.00. The app's activation event is logging a first bird sighting within seven days. On Network A, 300 of the 3,000 users log a sighting, so each active logger cost £10.00. On Network B, 600 of the 1,500 users log one, so each cost £5.00. Network B is twice as expensive per install and half as expensive per active user. The team moves budget from A to B and changes the campaign's optimisation event from install to first sighting.
What moves it
- The store listing. The product page in the app store is the step between the ad and the install; screenshots, ratings and the first lines of the description change how many visitors install.
- Audience and network. Broad or incentivised traffic installs readily and churns at once; narrow, interested audiences cost more per install.
- The optimisation event. Telling the platform to find installers produces cheap installs; telling it to find people who complete an in-app event produces dearer installs and better users.
- Creative. Ads that show what the app does attract people who want it; ads that promise something the app does not deliver inflate installs and deflate activation.
- Attribution limits. Privacy-limited measurement can undercount paid installs and so overstate CPI on some platforms, or credit installs to the wrong network.
Related metrics
CPI is the install-level case of CPACost of each conversion., and once installs are carried through to paying users the same spend becomes CACCost to acquire one new customer.. Whether any CPI is affordable is a question for LTVTotal value a customer brings over time.: the value a paying user returns, discounted by the share of installs that ever become one, sets the most an install can cost. The KPI tree for app installs lays these measures out as a chain from spend to retained user, so that a CPI change can be traced to the store page, the audience or the creative. How install campaigns are set up and optimised on particular platforms is covered in Google Ads for app installs and TikTok marketing for app installs.
What good looks like
There is no reliable universal benchmark, and the site does not cite one. CPI varies with platform, country, category, season and the optimisation event, and a published average says nothing about whether those installs open the app. The test that matters is internal: the cost per activated or paying user each network produces, compared over time, and whether that cost leaves room for the user to repay it from the value they generate before they leave.
Common mistakes
- Treating CPI as the result instead of the first step toward the cost of a useful user.
- Comparing CPI across networks with different attribution sources without noting which count each figure uses.
- Optimising to installs when the app earns money only from users who complete a later action.
- Mixing paid and organic installs in the denominator, which makes paid acquisition look cheaper.
- Ignoring delayed iOS reporting and judging a campaign before its installs have arrived.
- Leaving fees and production out of spend in one report and including them in another.
