Growth Engineering Playbook · Wave 5
Realised LTV always lies about your newest customers.
Lifetime value counted from banked orders (case 24) is honest but backward-looking — a cohort acquired last month has barely started paying, so its realised LTV looks terrible even if it's your best. This fits a retention curve to the cohort triangle, projects each cohort forward, and reports projected LTV next to realised so you can see how much is banked and how much is forecast. And it watches the one number that leads everything — early retention across cohorts — because that's where declining acquisition quality shows up long before LTV can confirm it. Deterministic, client-side.
Fitting retention curve…
Retention triangle
Active-customer share by cohort (rows) and months since acquisition (columns). Dashed cells are the future the projection fills.
Retention curve
Pooled observed retention (dots) and the fitted power-law curve extended into the projected tail.
Realised vs projected LTV by cohort
Green = value already banked; blue = value the projection expects still to come.
Cohort detail
Early retention (M1, M3) is the leading quality signal; realised vs projected shows how much of each cohort's value is still a forecast.