Cohort Analysis of Deposits
Cohort analysis of deposits shows what traffic actually earns over a defined period, not a blended daily average across the whole campaign. Here's how to build an install cohort and what it reveals about LTV that a standard daily report never shows.
What an install cohort is and why buyers need one
A cohort is a group of installs sharing a common trait, usually install date (or week), source and geo. Instead of asking how many deposits came in today, cohort analysis asks how much a specific group of installs ultimately generated over N days of its life. That distinction matters most where the deposit doesn't happen on install day but days or weeks later.
How to build an install cohort
Anchor the cohort on install date, lock in traffic source, creative and geo, then track what share of that cohort reaches registration and deposit at day 1, 3, 7 and 30 after install. Each new cohort is its own row in the table, never blended with previous days. The more precisely the cohort trait is defined, the more honest the comparison between one group and another.
What cohorts reveal that daily numbers hide
A daily report shows deposits that landed today — but those deposits may have come from installs two or three weeks ago, not from today's traffic. A cohort shows delayed deposits, day-of-week seasonality, creative fatigue (when later cohorts clearly underperform earlier ones), and gradual retention decay. Without cohorts it's easy to overrate today's traffic when the deposits showing up right now are actually being carried by an older cohort.
Estimating LTV from cohorts
Cohort LTV is the cumulative deposit total for one specific group of installs, measured over a long enough horizon — 30 or 60 days, not a forecast from the first 24 hours. Comparing a cohort's accumulated LTV against that same cohort's acquisition cost (CPI for the period) shows whether traffic pays back over time, even if the early picture looks weak. The estimate should stay qualitative — this cohort is paying back slower than the last one — rather than a fabricated precise forecast number.
Where the data for cohort analysis comes from
The foundation is accurate, timely postbacks — if a deposit event arrives late or gets duplicated, the cohort gets miscounted. APEX's S2S postbacks (Keitaro, FB CAPI and others) pass events without browser-side loss, which matters most for the delayed events cohort analysis depends on. Role-based access (buyer, team lead, CEO) is useful when rolled-up cohort dashboards need to reach the team without handing out raw tracker access.
FAQ
- How often should cohorts be recalculated?
- It makes sense to refresh daily for recent active cohorts, roughly their first 7 to 30 days, and weekly for older cohorts whose numbers have already settled.
- How many days should you wait before drawing LTV conclusions?
- It depends on vertical and typical deposit window, but conclusions drawn under a 7-day horizon are almost always premature. A 30-day window usually gives a far more reliable picture.
- How is cohort analysis different from a standard date-based report?
- A date-based report blends deposits from installs of different periods into a single number for today, while a cohort tracks one group of installs separately across its entire lifecycle. They answer different questions — how much came in today versus how much that specific install group ultimately generated.
- How do APEX postbacks help with cohort analysis?
- S2S deposit delivery without delays or duplicates matters most for cohorts, since a deposit can land days after install — a lost or duplicated postback specifically distorts that delayed portion of the numbers.
Cloaking, anti-bot, push, split tests and your own domains — in one service.
Get started free