What Retention does for you
Exact-day D1, D7, D14, and D30 return cohorts with mature denominators, lifecycle stages, and first-session friction.
Exact-day D1/D7/D14/D30
A player returning on day 10 does not count as a day-7 return. Each age has its own eligible denominator.
Mature cohorts only
Cohorts too young to observe a window are excluded from the percentage instead of depressing it.
Lifecycle stages
New, returning, at-risk, and lapsed players are grouped so the next action is visible, not just the curve.
Activity Index
Five engagement groups from inactive to very active, based on session frequency and playtime.
First-session friction
See how long new players stay in their first session and where onboarding loses them.
Retention by source
Compare return rates for each acquisition source to learn which campaigns bring players who stay.
How you'd use it
Three real situations owners and staff run into, and what retention answers.
Diagnosing a D7 drop
Day-7 return moved down two weeks in a row and nobody agrees on why.
Compare mature cohorts before and after the change, split by source and lifecycle, and separate a real drop from a young cohort.
Finding at-risk regulars
Your most engaged players are quietly playing less.
The lifecycle and Activity Index views surface players moving from active to irregular, with profiles one click away.
Choosing where to spend
Two campaigns bring similar join numbers but very different value.
Source-level retention shows which one produces players who return, so budget follows quality rather than volume.
How it works
Pick a cohort window
Choose the acquisition window and the return ages you care about.
Read mature denominators
Each percentage shows the eligible cohort behind it, so incomplete data cannot look like churn.
Act on a segment
Drill from a cohort into the players inside it and compare with campaigns or gameplay.
Questions
How do you define a day-7 return?
A player returns on the seventh calendar day after their first recorded join, in the reporting timezone. A return on day 10 is not counted as a day-7 return.
Why is a recent cohort shown as unavailable?
The cohort is too young to observe that return age. ShardAnalytics withholds the percentage instead of treating missing observation time as churn.
Is this a predictive churn model?
No. These are observed return and lifecycle metrics with rule-based groups. They show what happened, not a forecast.