← All tools
Analytics utility · Cohort segmentation

Blended retention mix analyzer.

An average is useful for reporting. It is not automatically the right number for allocation. Split the blend and see what moved inside it.

For each segment enter its share of installs and its retention, before and after. The default numbers reproduce the illustrative example from The Averages Trap.

Your numbers

Blended before—Weighted by share
Blended after——
Rate effect—Segments retaining differently
Mix effect—Shares moving

—

Change in retention by segment (points)
How it's calculated

The method, in plain terms.

  1. Blended retention = Σ (segment share × segment retention)
  2. Mix effect = change from shares at constant retention
  3. Rate effect = change from retention at constant shares
Worked example

A Tier-1 cohort at 20% share retaining 12% at D30 and the rest at 7% blend to 8%. If Tier-1 falls to 6% while the rest improves to 8.5%, the blend is still 8% — the decline is invisible in the average.

Defaults on this page are illustrative starting points, not benchmarks for your game. Replace them with your own numbers.

The thinking behind it
Terms on this page

Mix shift · Retention rate · Cohort

Embed this calculator

Free to embed on your blog, wiki or course page with the credit link intact.

FAQ

Common questions.

Why can blended retention stay flat while a market is getting worse?

Blended retention is a weighted average of segments. If one segment deteriorates while another improves, or the mix shifts toward a stronger segment, the average can stay put. The movement that matters for the next budget decision is hidden inside it.

What is the difference between a mix effect and a rate effect?

The rate effect is the change caused by segments retaining better or worse at constant shares. The mix effect is the change caused by segment shares moving at constant retention. Together they add up to the total change in the blended number.

Which segments should I split by?

Split by whatever receives the next decision: country or tier if the next UA dollar goes to a market, source or creative if it goes to a channel, platform or app version if the decision is technical. The MMP is usually where these splits live.

Does this work for ARPDAU or ROAS too?

Yes. Any blended ratio can be decomposed the same way; enter the metric values in the retention fields and the shares of the relevant base.

Calculator by The Game Scientist ↗