Playable Ad Metrics & Benchmarks: IPM, Interaction Rate & CPI by Genre

Playable performance is often boiled down to CPI, which hides the four other numbers that explain why CPI is where it is and which way it is heading. Read all five and you can find the actual problem instead of guessing which element to change.

The 5 Metrics That Matter

• IPM (Installs Per Mille): installs per 1,000 impressions. The primary volume signal. High IPM = the creative generates installs efficiently from the impression pool.
• Interaction Rate: percentage of impressions where the user engages with the mechanic (not just views). Signals creative engagement quality.
• Install Rate: of users who saw the CTA, what percentage tapped. Measures CTA effectiveness after engagement.
• CPI (Cost Per Install): total spend divided by installs. The cost efficiency metric. Driven by both IPM and CPM (what you're paying per thousand impressions).
• D1 Retention Rate: percentage of users still active 24 hours after install. Measures install quality from the creative's pre-qualification effectiveness.

Industry Benchmark Ranges

These ranges are directional: actual benchmarks vary by network, geography, device mix, and genre maturity. Use them to calibrate whether a metric is 'in the ballpark' rather than as hard targets.

• Casual game IPM (interstitial placement): 12–30. Hyper-casual with simple end card: can exceed 40.
• Mid-core / strategy game IPM: 4–12. Lower because the audience is narrower.
• Interaction rate (simple tap mechanic): 18–35%. Rate drops as mechanic complexity increases.
• Interaction rate (hold or path mechanic): 8–20%. The friction filters for more committed users.
• CPI (casual, tier-1 markets, 2026): $1.50–$4.00 for playables. Video CPI is typically $2.00–$6.00 for the same audience.
• D1 retention benchmark for playable-acquired users: 30–45% (casual), 20–35% (mid-core). If below these, the creative may be attracting the wrong audience.

CPI Is a Product of CPM and IPM

CPI = CPM / IPM × 1000. A CPI of $3.00 on Google UAC might come from a $30 CPM and an IPM of 10, or from a $15 CPM and an IPM of 5. These require different fixes. High CPM with acceptable IPM: the creative is working but you're paying a premium for the audience segment (normal for tier-1 geos). Low IPM with normal CPM: the creative isn't converting impressions (the mechanic, headline, or CTA is underperforming). Separate these before deciding what to change.

Diagnosing Low Interaction Rate

Low interaction rate (under 10% for a simple tap mechanic) usually means one of three things: the interaction prompt is visually unclear (users don't see it or don't understand what to tap), the interaction appears too late (after 5+ seconds, when users have scrolled), or the mechanic requires reading instructions (friction kills spontaneous interaction). Fix in that order: make the prompt bigger and more obvious, move it earlier in the creative timeline, simplify the mechanic to a single tap. See game genre creative strategy for genre-specific interaction design.

Reading One Creative as a Funnel

The five metrics above are more useful read in sequence than one at a time, because each stage tells you where the next one lost people. The figures below are an illustration of the arithmetic only, not targets. Say one creative served 20,000 impressions on one network in a week:

1. Impressions: 20,000.
2. Interactions: 3,000 people touched the mechanic (an interaction rate of 15%).
3. Round completed: 1,800 of those reached the end of the round (60% of interactors).
4. CTA taps: 600 tapped through to the store.
5. Installs: 200 (an IPM of 10).
6. Still active the next day: 70 (a D1 retention of 35% for this cohort).

Now each drop is a specific question. If interactions are low, the prompt or the first second is the problem. If many people interact but few finish, the round is too long or too hard. If people finish but few tap through, the end card is weak. If taps are healthy but installs are not, the store page is losing them, which no creative change will fix. And if installs are healthy but retention is poor, the creative is promising something the game does not deliver.

Where Each Number Comes From

No single dashboard holds the whole funnel, and mixing sources without knowing it is a common source of contradictions.

• Impressions, spend, CPM and network-reported installs come from the network's own reporting.
• Interaction, completion and CTA events come from inside the creative, and only where the network supports and surfaces them. See playable ad analytics and event tracking.
• Attributed installs, retention and revenue per creative come from your MMP, which applies its own attribution rules.
• On iOS, much of this arrives through SKAdNetwork, with less detail and a delay. See SKAdNetwork and creative measurement.

Network installs and MMP installs will rarely match exactly. Pick one source per metric, write down which, and compare like with like.

When to Stop a Test

Stop a creative test when: (1) one variant has 2× the IPM of the other AND 500+ impressions per variant (early strong signal), or (2) both variants have 5,000+ impressions and the IPM difference is under 10% (no meaningful difference: test another variable). Don't stop tests during the first 48 hours: ML bidding systems deprioritize new creatives initially, which artificially depresses IPM before the system calibrates. See A/B testing methodology for full stopping criteria.

Comparing Across Networks and Placements

An IPM on one network and an IPM on another are not the same measurement of the same thing. Networks count impressions and attribute installs differently, serve different audiences, and run different auctions. Placement matters as much: a rewarded placement, where the user chose to watch, and an interstitial, where they did not, produce different engagement for the identical file (see rewarded vs interstitial playables). Compare creatives within one network and placement; compare networks only on the metric you actually pay for, measured in one place.

Signs a Metric Is Misleading You

• A dramatic result on a small sample. A creative with a handful of installs can show any IPM at all.
• A new creative judged during the delivery system's learning period. See the learning phase.
• A slow decline that looks like a creative problem but is fatigue: the same people seeing it again. See creative fatigue.
• A sudden change on the day an attribution setting, SDK version or tracking rule changed. The creative did not move; the measurement did.
• Strong install numbers with retention that collapses on day one, which can point to low-quality or accidental traffic rather than a great creative.

When a number surprises you, find the condition behind it before acting on it. A metric that is true only under one condition (one network, one placement, one week) is a narrower claim than it looks on the dashboard.