Originally published November 2019. The original Pollen VC page is unavailable. The archive we could recover preserves its title, publication metadata and opening, rather than the complete article.
Short answer: a projected lifetime value (LTV) is a forecast. Backtesting compares a model’s prediction with the revenue the same acquired cohort actually produced over the corresponding period. Use that comparison to reassess future spending assumptions.
I wrote this piece for Pollen VC in November 2019. The archived opening describes a familiar setup: you have revenue, spend on paid acquisition and a projected LTV that suggests acquisition is profitable. What could go wrong?
Our original post described three reasons actual revenue can diverge from projected revenue. Those reasons were not recoverable from the archive we checked, so this page preserves the resource’s premise without reconstructing its missing argument.
How does backtesting inform a spending decision?
Compare forecasts and realized revenue for equivalent cohort ages. A projection of a full year’s revenue cannot be judged against only a month of receipts. Repeated gaps are a reason to revisit assumptions before using the model to justify more spending. This is general modeling guidance, not a reconstruction of the missing article.
For related resources, see calculating LTV without a lifetime of data, why LTV still needs cash and ROAS for mobile apps.
