The problem with averages
Aggregates blend together customers acquired at different times, prices, and behaviors. When your product or acquisition channel changes, averages can look stable while underlying cohorts deteriorate (or improve).
What is a cohort-
A cohort is a group of customers that share a start date (e.g., customers acquired in January). Cohort analysis tracks retention, expansion, and churn for that group over time.
When cohort analysis matters most
- When you change pricing, packaging, or onboarding.
- When you add a new acquisition channel or targeting strategy.
- When you're trying to forecast LTV or payback more accurately.
Common aggregate traps
- Blended churn hides a weak new cohort even as older cohorts remain stable.
- Expansion from a few large accounts can mask broad-based churn.
- Promo-heavy cohorts can inflate short-term revenue and deflate true LTV.
Cohort QA checklist
- Use a consistent cohort start event (signup or first value).
- Track cohorts by acquisition channel and plan for clarity.
- Normalize for billing cadence (monthly vs annual) when comparing retention.
How to use cohorts with LTV and payback
- Compute retention/churn by cohort and segment (plan, channel, geo).
- Estimate LTV from retention curves rather than a single churn number.
- Compare payback by cohort to see whether acquisition quality is changing.
When aggregates are still useful
- For quick executive dashboards where trend direction is the goal.
- When cohort data is too sparse to be statistically useful.
- For short-term operational monitoring (daily or weekly changes).