What is Cost Per Install (CPI)?
Also known as: CPI
What is CPI?
Cost per install (CPI) is total ad spend divided by attributed app installs over a period. It is the core volume metric in mobile user acquisition. Per AppsFlyer's Performance Index 2024, global app install ad spend topped $94 billion in 2023, with CPI as the primary cross-channel rate.
CPI answers one question. How much did one install cost. It does not tell you if that install was worth buying. That part requires retention, ARPU, and payback math against app monetization reality.
Networks report CPI on a campaign, ad set, geo, creative, or device basis. UA managers read it daily. Finance teams read it monthly.
CPI formula and example
The formula is simple. Total media spend on a campaign divided by attributed installs from that campaign.
CPI = Ad spend / Attributed installsA casual game spends $30,000 on a Meta Advantage+ App Campaign. The campaign drives 12,500 installs. CPI is $2.40. Run the same math per geo and the picture fragments. US installs cost $4.10. Brazil installs cost $0.85. India installs cost $0.42.
The metric is only as clean as the attribution model behind it. SKAN postbacks, MMP probabilistic matching, and self-attributing networks each count installs differently. Reconcile before reporting.
Average CPI by platform and geo
CPI varies wildly by OS, geo, and category. The numbers below combine AppsFlyer's Performance Index 2024 and Adjust's 2024 mobile benchmarks for casual gaming and non-gaming apps.
| Platform / Geo | Casual games CPI | Non-gaming apps CPI |
|---|---|---|
| iOS, US | $4.20 | $6.80 |
| iOS, Western Europe | $3.10 | $5.40 |
| Android, US | $1.80 | $3.90 |
| Android, Western Europe | $1.30 | $2.80 |
| Android, LATAM | $0.85 | $1.60 |
| Android, India | $0.40 | $0.90 |
iOS sits 2 to 4 times above Android in the same geo. Tier 1 geos sit 4 to 10 times above tier 3. Subscription apps and fintech sit at the top of the curve. Hyper-casual games sit at the bottom.
CPI vs CPA vs CPM in mobile UA
Three rate metrics get used together. They measure different parts of the funnel.
| Metric | What it counts | Where it sits | Typical range |
|---|---|---|---|
| CPM | Cost per 1,000 impressions | Top of funnel | $1 to $25 |
| CPI | Cost per install | Mid funnel | $0.40 to $8 |
| CPA | Cost per acquisition event (signup, trial, purchase) | Bottom funnel | $5 to $80 |
CPM tells you how expensive the auction is. CPI tells you how well your store listing converts paid traffic. Cost per acquisition tells you how well your onboarding turns installs into real users. The three move together, but not always in the same direction.
What drives CPI up or down
Five levers move CPI. Most teams adjust them in this order.
- Creative. A new winning video can cut CPI by 30 to 50 percent overnight. Per AppsFlyer's Performance Index 2024, accounts that ship 5 or more new creatives per week run 28 percent below category median CPI.
- Geo mix. Shifting budget from US to LATAM or India drops blended CPI fast. It also drops LTV. Net economics matter, not the rate alone.
- Bidding model. Manual CPI bidding is mostly gone. tCPA and tROAS bidding raises the apparent CPI but lifts cohort quality. The trade is almost always worth it.
- Store listing. A/B tested icons, screenshots, and preview videos lift install conversion rate. Same impressions, more installs, lower CPI.
- Audience saturation. As a campaign scales, the auction price climbs. CPI drift up by 10 to 20 percent at scale is normal.
Post-ATT measurement
iOS CPI reporting changed permanently in April 2021. Apple's App Tracking Transparency requires user opt-in for the IDFA. Per Adjust's 2024 mobile benchmarks, iOS opt-in plateaued between 25 and 38 percent across categories.
UA teams now blend three signals to read iOS CPI. SKAdNetwork postbacks from Apple, deterministic data from opted-in users, and probabilistic models from MMPs like AppsFlyer and Adjust. SKAN 4 added more conversion windows and coarse value tiers, but no user-level data.
The practical effect on CPI reporting is a 24 to 72 hour lag and a 5 to 15 percent variance between ad platform and MMP numbers. Reconcile weekly, not daily.
Real-world example
A mid-core RPG launches in 5 geos with a $150,000 monthly budget. Day-7 retention is 18 percent. ARPDAU is $0.42.
| Geo | Spend | Installs | CPI | D7 retention | 90-day LTV |
|---|---|---|---|---|---|
| US (iOS) | $50,000 | 11,900 | $4.20 | 22% | $9.10 |
| US (Android) | $40,000 | 22,200 | $1.80 | 19% | $5.80 |
| Germany (Android) | $25,000 | 19,200 | $1.30 | 17% | $4.20 |
| Brazil (Android) | $20,000 | 23,500 | $0.85 | 14% | $1.90 |
| India (Android) | $15,000 | 37,500 | $0.40 | 11% | $0.60 |
India delivers the lowest CPI by a factor of 10 against US iOS. It also delivers the worst LTV-to-CPI ratio at 1.5 to 1. US iOS sits at 2.2 to 1. Germany Android leads at 3.2 to 1.
The UA manager cuts India by half, holds Brazil flat, and pushes incremental budget into Germany Android and US iOS. Blended CPI rises from $1.43 to $2.10. Blended LTV rises from $3.20 to $5.40. The campaign moves from break-even to a 2.6 to 1 LTV-to-CPI ratio without raising total spend.
CPI alone said the opposite. That is why no serious UA team optimizes against it directly.
Related terms
Frequently asked questions
What is a good CPI for mobile apps?
It depends on platform and category. Per AppsFlyer's Performance Index 2024, median Android CPI for casual games sits near $1.20, while iOS subscription apps run $4 to $7. The right CPI is whatever lets payback land inside 12 months.
Is CPI the same as CPA?
No. CPI counts every install. Cost per acquisition counts a deeper event, like a signup, trial, or first purchase. CPI is always lower than CPA. Most teams now bid on CPA or ROAS and treat CPI as a diagnostic, not a target.
How did iOS 14.5 ATT change CPI?
App Tracking Transparency broke deterministic iOS attribution after April 2021. Per Adjust's 2024 mobile benchmarks, iOS opt-in sits between 25 and 38 percent. iOS CPIs dropped on paper as advertisers fled, then rebounded. SKAdNetwork postbacks now carry most iOS CPI reporting.
Why is my Android CPI lower than iOS?
Three reasons. Android has more global inventory, especially in tier 2 and tier 3 geos. iOS users skew higher income and ad networks bid harder for them. iOS attribution is also noisier post-ATT, so platforms underreport installs. Expect iOS CPI to run 2 to 4 times Android in the same category.
Should I optimize campaigns toward CPI?
Rarely. CPI bidding rewards cheap, low-quality installs. Most UA teams now bid on tCPA, tROAS, or value-based events. Use CPI as a rate check across creative and geo, not as the optimization target. The exception is hyper-casual games where ad LTV is flat across users.