Mobile App CPI Benchmarks 2026: By Platform, Country & Category

"Is our CPI good?" is the question every founder asks and every benchmark report answers badly—usually with a single global average that is meaningless for your app. Real cost per install varies sharply by country, platform, category, and channel, and the blend you should expect depends on which combination you are buying. This post lays out the working ranges we see across live consumer-app accounts in 2026, why those ranges look the way they do, and—the part most benchmark posts skip—how to use a benchmark without letting it wreck your strategy.

Methodology, said plainly: these are working ranges from accounts we run or audit, cross-checked against public industry compilations such as Business of Apps' CPI research. They describe broad-targeted, conversion-optimised campaigns at meaningful spend—not hyper-niche audiences, not pure retargeting, not a $500 learning budget. Treat every figure as a range with a reason. For an interactive sanity check against category bands, use our free benchmarks tool.

How to read a CPI number without fooling yourself

CPI is a channel metric, not a business metric. Two apps can share a $3.00 CPI and have opposite health: one pays back in ninety days, the other never. Before you celebrate or panic, put CPI next to cost per paying user, early ROAS, and payback month. Our payback period guide and ROAS calculator exist specifically so benchmarks stay in the "context" column instead of the "decision" column.

Also separate blended CPI from marginal CPI. Blended includes cheap leftovers and brand. Scaling decisions happen at the margin—what the next thousand installs cost—which is almost always higher than the average that makes dashboards look calm.

By platform: iOS costs more, and it is often worth it

In 2026, iOS CPI commonly runs materially above Android for the same app in the same market—often on the order of roughly a third to half higher in competitive consumer categories, sometimes more. Three forces keep the gap open: stronger monetisation draws more advertisers into iOS auctions, Apple's privacy constraints make optimisation less efficient (raising effective prices), and iPhone demographics in tier-1 markets skew toward higher ability to pay.

The caveat that matters: iOS users still often monetise meaningfully better than Android users in subscription categories. A higher iOS CPI can coexist with a lower cost per payer. If you reallocate on CPI alone you will systematically underinvest in iOS. Pull the comparison to trial starts, subscriptions, and payback before you "fix" the platform mix.

By market: think in tiers, not trivia

1. 01

   Tier 1 (US, UK, Canada, Australia)

   the most expensive auctions. Mainstream consumer apps often see multi-dollar CPIs; dating and fintech sit higher still. The US usually clears above the UK for the same creative.

2. 02

   Western Europe (DACH, Nordics, France, Benelux)

   commonly cheaper than the US with solid subscription monetisation—chronically underrated by US-centric teams.

3. 03

   Tier 2 (Eastern Europe, LatAm, Southeast Asia)

   sub-dollar to low-dollar installs are common. Monetisation is lower, but these markets are excellent creative labs: same hook-rate signals at a fraction of tier-1 cost.

4. 04

   Emerging (India, parts of Africa/MENA)

   very cheap installs. Only useful if monetisation works there—ad-monetised products often do; hard-paywall subscriptions often do not.

The testing-lab pattern is one of the highest-ROI habits available: run new creative batches in cheaper markets first, kill losers early, promote only proven winners into the US auction. Creative signal transfers across markets better than most teams assume. Structure those tests with our 60-day creative testing framework.

By category: competition sets the price

Category bands move with advertiser density and user value, not with your feelings about your product:

1. 01

   Casual / hyper-casual games

   typically the cheaper end—broad appeal, liquid auctions.

2. 02

   Photo, video, utility

   moderate CPIs, high volume potential.

3. 03

   Health & fitness

   mid bands, with seasonal spikes (January is real).

4. 04

   Subscription lifestyle (sleep, journaling, language, meditation)

   mid bands in tier-1; creative quality dominates variance.

5. 05

   Dating

   expensive and defended.

6. 06

   Fintech & investing

   top of the chart—you compete with deep-pocketed financial advertisers and often need higher-intent users than a casual install.

If your store category is awkward, pick an analogue by monetisation model and decision weight. An app that monetises like fintech will often price like fintech even if it lives under Productivity. Use the midpoint of a public table only as a prior, then manage your own eight-week trend.

By channel: where the relative discounts live

TikTok frequently delivers lower CPIs than Meta for the same app when inventory outpaces advertiser demand—especially outside the hottest US auctions—at the cost of somewhat noisier user quality for narrow, high-LTV products. Meta remains the workhorse for event optimisation when you feed it a clean down-funnel signal; inspect competitors' angles in the Meta Ad Library and creative patterns in TikTok Creative Center before you assume a channel is "expensive" rather than creatively tired.

Apple Search Ads is the inverse of broad social: high intent, high price, excellent for brand defence and high-converting discovery queries—not your sole scale engine. Google App campaigns usually sit between social channels on price with less creative control. Use Google Ads Transparency Center to see how rivals show up on YouTube and Search when you benchmark message, not just cost.

The benchmark trap

A benchmark answers "is our market position sane?" Teams keep using it to answer "are we doing well?" That confusion creates two expensive failure modes:

1. 01

   Cheap-install celebration

   Below-benchmark CPI with weak D30 retention is a subsidised leak. You are paying less to lose money.

2. 02

   Expensive-install panic

   Above-benchmark CPI with strong retention and financeable payback is a machine that deserves more budget. Some of the best accounts we run intentionally sit in the expensive quartile because creative targets payers, not browsers.

Decide with blended CPI trend (falling while volume grows is healthy), cost per payer by cohort, and payback window. For many consumer apps, a few months to roughly half a year of payback is the financeable zone. Benchmarks are context; cohort economics are the decision. Translate your CPI ceiling into a creative rule with the D7 ROAS kill line.

What actually moves CPI (hint: not bids)

Across accounts we run, the gap between an average month and a great month is rarely bid theatre. It is creative velocity: genuinely new concepts entering tests. Platforms allocate cheap distribution to fresh creatives with strong early engagement because engaging ads make the platform money. Winners get a discount; fatigued assets pay a compounding tax that shows up as mysterious CPI creep even when "targeting did not change."

Teams shipping a high volume of distinct concepts—not forty resizes of one video—consistently sit below their category's painful edge. Kill losers fast, scale winners hard, and keep production cheap enough that weekly variants are normal. Tactics live in how to lower CPI in 2026; economics live in the Payback Engine.

A practical way to use this post next week

Pick your platform × tier × category band. Plot your last eight weeks of marginal CPI against it. If you are inside the band and payback is on target, stop optimising CPI for sport and invest in volume. If you are above band and payback is slipping, audit creative age and concept diversity before you touch bids. If you are below band with soft retention, raise targeting or creative specificity—even if CPI rises.

Also write down the reason you believe your CPI should differ from the band: niche audience, web-to-app funnel, heavy brand, or a creative advantage. If you cannot name a reason, assume you are average and manage accordingly. Average is not an insult; unexamined variance is.

Exp(G) operating ranges vs published tables

When we quote CPI bands in client reviews, we label them as Exp(G) operating ranges — portfolio experience from accounts we run or audit — not audited industry averages. Published tables from Business of Apps, MMP reports, and store-intelligence vendors will disagree; sample mix and SKAN modelling differ. Use external tables for neighbourhood checks. Use your Payback Engine output for scale decisions. Before throwing budget at a geo that only looks cheap because the hook is tired, score the opening claim in the hook analyzer and re-check break-even in the ROAS calculator. Cheap inventory with a dead hook is still expensive.

If your CPI has crept above the ranges that make sense for your mix and bid changes are not fixing it, the leak is usually creative or monetisation timing—not a missing bid modifier. That production-and-testing system is what we build inside client engagements, with a result guarantee on the KPI you choose. Book a discovery call and bring your last sixty days of CPI by channel; we will show you which lever moves first.

Sources & further reading

• Business of Apps — CPI research — public industry compilations for cross-checks
• Meta Ad Library — see what you are bidding against creatively
• TikTok Creative Center — creative benchmarks and top ads by market
• Google Ads Transparency Center — Search/YouTube competitive presence
• Apple Search Ads — high-intent iOS acquisition and keyword popularity signals
• Sensor Tower — market context around category competition
• AppMagic — accessible download and creative market intel