A cheap install is not a cheap user
Every ad network shows CPI — cost per install — and every dashboard makes a low number look like a win. It isn't. An install is only worth what the player goes on to spend, and with day-1 retention around 27%, day-7 around 13%, and day-30 around 5% (Adjust, 2025), most installs never become revenue. The real question is not "how cheap was the install" but "did the cohort pay back."
What CPI actually looks like in 2025
CPI swings wildly by platform and market — any single "average" hides the most important variable. Adjust's 2026 report puts the global gaming CPI at $0.56 (2025, +30% year over year), but that floor is dominated by low-cost, high-volume geos. Tier-one markets pay far more, and iOS runs well above Android inside the same genre.
| Market / split | 2025 CPI | Note |
|---|---|---|
| Global gaming (Adjust) | $0.56 | +30% YoY — low-cost geo floor |
| North America | $1.68 | US $1.71 (+31%) |
| Europe | $0.53 | DACH $1.70 |
| APAC | $0.27 | Singapore $2.49 |
| LATAM | $0.14 | cheapest for early testing |
| iOS vs Android casual (Liftoff) | $1.41 vs $0.14 | ~10× spread in one genre |
Liftoff's 2025 casual report measured casual CPI at $1.41 on iOS against $0.14 on Android — a tenfold gap. Singular's Q4 2025 data points the same way: Android game CPIs typically run three to four times below iOS. A benchmark that doesn't split platform and market is averaging away the decision that matters.
ROAS is the metric that decides scale
Return on ad spend is measured at a named window because revenue accrues over weeks. Liftoff's 2025 Mobile Gaming Apps Report gives the genre baselines: ad-led casual titles recover fast, IAP-heavy titles take months.
| Model | D7 ROAS | D30 ROAS | Payback |
|---|---|---|---|
| Ad-led (casual / hyper) | ~7.6% | ~47% | ~30 days |
| Mid-core RPG / strategy | ~4.3% | 30–40% | 90–180 days |
| IAP-heavy F2P (target) | 15–40% | 60%+ | 90–180 days |
The trap is applying one ROAS threshold across every model. A D7 of 7–8% is on track for an ad-monetized casual game if D30 reaches ~47%; the same number would be a disaster for an IAP title. The accepted rule: keep CPI at 30–70% of projected LTV, and set the payback window by how the game earns.
Why a runtime makes the math cleaner
Mini-games rarely live in one store. They run inside several host super-apps, each with its own audience and its own traffic. The risk is that UA becomes a per-host silo: you compare ROAS in store A against store B and never see the blended picture. A shared runtime flips that — one cohort definition, one attribution view, one dashboard across every host the game ships to. You still buy installs where they make sense, but you measure payback on the player, not the store.
The takeaway
Treat CPI as an input, not a verdict. Split it by platform and market, track D7 and D30 ROAS against the right baseline for your model, and keep CPI under 70% of LTV. With more than 72% of mobile developers now running a combined IAP + ad model (AppsFlyer, H1 2025), the games that win are the ones that blend paid acquisition with organic reach from the host — and measure both on one canvas.
Cross Mini App gives every mini-game one runtime and one cohort view across all host apps, so UA decisions are made on the player, not fragmented across stores.
