1. What it means
- Cohort analysis = group customers by when they first joined (e.g., Jan 2025 cohort, Feb 2025 cohort).
- Cohort-based LTV = track how much revenue each cohort generates over time → average it → estimate lifetime value.
Instead of assuming churn is constant, we actually observe how cohorts behave month by month.
2. Steps (simple method)
- Define cohorts
Example: all customers acquired in January 2025 = one cohort. - Track revenue (or gross profit) per cohort, over time
For each month after signup, calculate average revenue per customer in that cohort. Example (per-customer revenue):- Month 0: \$100 (signup fee)
- Month 1: \$40
- Month 2: \$38
- Month 3: \$35
…
- Calculate cumulative revenue
Add up average revenue across months until the cohort stabilizes (or churns out). Example:- By Month 3: \$100 + 40 + 38 + 35 = $213
- By Month 6: \$100 + 40 + 38 + 35 + 32 + 30 + 28 = $273
- Estimate “lifetime”
If after ~12 months the curve flattens (most churned), take the total.
If not, fit a simple decay model (exponential/linear) to extend beyond observed months.
3. Simple Formula (if you use average per customer over cohorts)
$\text{LTV} = \sum_{t=0}^{T} \frac{\text{Avg Revenue per Customer in Month } t}{(1+r)^t}$
- $T$ = number of months you track
- $r$ = discount rate (optional, for NPV adjustment; often ignored in simple calc).
4. Numerical Example (no discounting)
- Cohort size = 100 new customers in Jan 2025.
- Revenue per customer over first 6 months (average):
| Month | Avg Rev/Customer |
|---|---|
| 0 | $100 |
| 1 | $40 |
| 2 | $38 |
| 3 | $35 |
| 4 | $32 |
| 5 | $30 |
LTV (6 months) = 100 + 40 + 38 + 35 + 32 + 30 = $275 per customer
If revenue keeps declining, you can project future months with a decay assumption.
5. Advantages of cohort-based LTV
- Uses real observed behavior, not just churn assumptions.
- Highlights differences by acquisition month, channel, or segment.
- Helps see whether retention is improving or worsening across cohorts.
Summary:
Simple cohort-based LTV = sum of average revenue per customer over time for a given cohort.
It’s more accurate than the churn-based shortcut ($1/\text{churn}$) because it uses actual cohort retention and spend data.