Ad Frequency and Churn: Finding the Line

Published guidance converges on one to two interstitials per session, a minimum interval of two to three minutes between them, and a first-advert delay of 60 to 90 seconds. That delay alone is reported to improve day one retention by several points with little revenue loss. Rewarded video behaves differently and is not subject to the same limits.

Why this matters

Ad frequency is the fastest way to increase revenue per player and the fastest way to lose them.

The damage is also delayed and disguised. Raising frequency lifts revenue immediately and reduces sessions gradually, so the loss appears weeks later as a retention problem rather than as a monetisation decision.

Knowing the published starting points saves you from discovering the line by crossing it.

What frequency does to retention

The relationship is not simple, and one published finding is worth stating carefully.

Analysis of churn modelling has found interstitial impressions correlating negatively with churn, at around -0.48 in one published model. Players who saw more adverts were less likely to leave.

That is not evidence that adverts improve retention. It is a selection effect: engaged players play more, and playing more produces more impressions. Ad exposure is acting as a proxy for engagement.

The correct reading is that you cannot evaluate frequency by correlating impressions with churn across players. You have to change the frequency and compare groups.

Players who see more ads churn less, because they were playing more. That is a measurement trap, not a design finding.

Published timing guidance

Where the public sources agree.

Sources: AdReact's interstitial best practice analysis and Meegle's interstitial monetisation overview.

The first-advert delay is the highest-value item on that list. It is a configuration change, it costs no development time, and the reported retention effect is meaningful at a stage where every point of day one matters.

The curve, by placement type

Different placements carry very different tolerance.

Interstitials are the sensitive ones, because they interrupt. Tolerance depends heavily on where they sit: between levels is accepted, mid-level is not.

Banners are tolerated at much higher exposure, since they do not interrupt, though they cost screen space and can be tapped accidentally.

Rewarded video is the exception and is covered below.

App open adverts are the most damaging per impression, because they sit between the player and the thing they came for. Treat any use of them as an experiment rather than a default.

Interstitial timing rules that work

Beyond the published intervals, five placement rules matter as much as the frequency number.

1. Only at clean breaks. After a level completes or a run ends, never during play.
2. Never after a failure. A player who just lost and is then shown an advert is being charged for failing, and it is the most reliable way to end a session.
3. Never on the first session. Let the player understand what the game is before monetising them.
4. Never twice in succession. Two levels in a row with an advert reads as constant, whatever the average says.
5. Never during a streak. Interrupting an accumulating run costs more than the impression is worth.

Rule two is broken frequently and it is the most damaging. Failure is already a negative moment; adding an advert compounds it at exactly the point where the player is deciding whether to continue.

Rewarded video as the exception

Rewarded video does not follow the same curve, because the player chooses it.

That changes the economics. Published guidance on interstitial limits does not apply, and higher exposure is generally tolerated or welcomed.

Three rules keep it working.

• Always optional and clearly labelled, with the reward stated before the advert plays.
• Placed at a moment of specific want, such as a failure the player can recover from, or a capacity limit.
• Capped per session, so it does not become the main loop and displace the game.

Rewarded video also does the teaching work described in Hybrid-Casual Monetization. A player who has watched an advert for a hint understands exactly what a hint is worth, which is what makes a later purchase possible.

Testing frequency safely

The only reliable method is a controlled comparison, and it requires remote configuration.

1. Split new installs into groups, differing only in ad frequency.
2. Change one variable. Interval, or count per session, not both.
3. Run long enough to see day seven, since the loss shows up after the gain.
4. Compare revenue per player over the window, not revenue per session. Frequency raises the second while lowering the first.
5. Watch sessions per day, which moves before retention does and gives you an early warning.
6. Set a stopping rule so a group that is clearly losing players is ended early.

The fourth point is where most teams go wrong. Revenue per session rises with frequency almost by definition, and reading it as success is how a game ends up over-monetised.

Remote configuration is what makes this possible without a build, which is why it is the first backend feature worth having, per Owning Your Backend.

Where these numbers came from

All published, and all worth verifying against current sources.

• Interstitial frequency, interval, and first-advert delay guidance, plus the reported D1 effect: AdReact.
• General interstitial monetisation practice: Meegle.
• The impressions-to-churn correlation and its interpretation: a published churn prediction model.

These are starting points rather than answers. The right frequency depends on your session shape, your genre, and your markets, and only a controlled test on your own game will find it.

What we would do

Start conservative: one interstitial per session, a two to three minute minimum interval, no advert in the first ninety seconds, and none on the first session. Published guidance supports all four and they cost nothing to implement.

Put every one of those values in remote configuration before launch, so frequency can be tested without a build.

Then run one controlled test at a time and judge it on revenue per player over a week, with sessions per day as the early warning. The number you find will be specific to your game, and it will be worth considerably more than any published figure.

The short version

• Published guidance: one to two interstitials per session, two to three minutes apart, none in the first 60 to 90 seconds.
• The first-advert delay is a configuration change with a reported multi-point D1 improvement.
• Impressions correlate negatively with churn because engaged players see more; that is a trap, not a finding.
• Never show an advert after a failure, during a streak, or in the first session.
• Rewarded video is chosen by the player and does not follow the same limits.
• Test with remote configuration and judge on revenue per player, never per session.

Move your ad frequency values into remote configuration before launch. If you want a monetisation pass that does not cost retention, tell us about the game.

Related reading: Hybrid-Casual Monetization, How Mobile Games Make Money, and Session Length Versus Revenue.