Predicted LTV: fund and kill channels off the early signal
In 1693, an astronomer priced insurance for people he had never met, in a city he had never visited.
Edmond Halley. Yes, the comet one. He got hold of birth and death records from the city of Breslau and built one of the first real life tables from actual population data. It became a basis for pricing annuities. Not after someone died. Not after 12 months of "let's wait and see." Off the early pattern, priced on day one.
The mechanism
Halley was not guessing. He read a predicted future value off an early signal and acted on it before the real outcome arrived. That is the whole business model of life insurance. Wait for the realized number and there is no product left to sell.
Most growth teams do the opposite with LTV. A "final" 12-month number is trusted like scripture. A day-7 or day-30 predicted number is treated like a guess to hedge around.
The data says that is backwards
RevenueCat's data shows that for products with 2,000 or more paying subscribers, over 75% get 12-month LTV predictions that land within 10% of the real outcome. And the signal driving that accuracy shows up in the first 7 to 30 days. Not in month 12.
What waiting really costs
A growth team that waits for month 12 to price its CAC is not being careful. It is pricing the policy after the funeral.
By the time the real number lands, the budget decision that mattered has already been made, on gut feel, by whoever did not trust the math.
Where predictions go wrong
Predicted LTV is not magic. A prediction inherits the curve it was fit to. If your early cohorts had not yet hit a content ceiling or a pricing change, the model will carry that blind spot forward. Two habits keep it honest:
- Compare predictions with actuals every month, cohort by cohort, and refit when they drift.
- Never use an industry benchmark as the comparison point. Fit the curve to your own data.
The move
- Pick your early signal: the day 7 or day 30 revenue and retention numbers that best predicted 12-month value in your past cohorts.
- Fit a simple curve from those points to a 12-month LTV.
- Set your CAC ceiling from predicted LTV, with a margin of safety you agree on in advance.
- Fund the channel off the early signal. Kill it off the early signal too.
- Check predictions against real outcomes monthly and refit.
Bottom line
An insurer never waits for the death to price the policy. Your budget should not wait for month 12 either.
FAQ
What is predicted LTV?
Predicted lifetime value is an estimate of how much revenue a user or cohort will generate over a set period, such as 12 months, based on early behaviour like day 7 or day 30 revenue and retention.
How accurate is predicted LTV?
Volume and model quality decide it. RevenueCat reports that for products with at least 2,000 paying subscribers, over 75% get 12-month predictions within 10% of the actual result.
How early can I predict LTV for a subscription app?
Most of the useful signal appears in the first 7 to 30 days. That is usually enough to make budget decisions, as long as you keep checking predictions against actuals.
Should I compare my LTV to industry benchmarks?
No. Benchmarks describe other businesses. Fit the curve to your own cohorts, so the prediction reflects how your users actually behave.