Retention moved. Which lever did it?
UA mix, onboarding, core game: when they all changed at once, dashboards can’t tell you which one moved the number. Asemic’s causal decomposition splits the change into its drivers, so you fix the right thing.
Be first in line.
Get the launch announcement and the case studies as we publish them: how real retention moves decompose in practice. No spam, unsubscribe anytime.
Running a studio and want in early? Book a demo. We’re onboarding design partners now.
Attribution for product decisions,
not just ad clicks.
Compare the world as it was
Two periods or two cohorts, resolved point-in-time: who you acquired, what their early days looked like, what was live in the game. Apples to apples, as of the moment it happened.
Attribute the change
The engine splits the KPI delta into contributions: how much came from who you brought in (UA), how much from the early experience (onboarding), and how much from what you shipped (core-game).
Act on the driver
A −2.8pp retention move stops being a mystery meeting and becomes a decision: fix the channel mix, the FTUE, or the patch, whichever actually moved it.
The decomposition is only as trustworthy as its cohorts, which is why it sits on Asemic’s point-in-time engine: every comparison resolves player state as of the day it describes.
Validated against known ground truth.
Causal claims are easy to make and hard to earn. Ours are tested three ways:
Real games
Tested on event data from multiple live titles, not synthetic demos.
A/B ground truth
Checked against experiments where the true effect is known from the test itself.
Injected effects
Run on datasets with deliberately planted changes; the algorithm recovers the known answer.
Honest fine print: known edge cases are still being refined; we say “validated against ground truth,” never “perfect.” Retention decomposition ships first; revenue is next.
Stop guessing which lever moved the metric.
Launch news, plus the case studies as they’re published.