Campaign LTV Modeling and Spend Recovery

Blinkit is a hyperlocal quick commerce platform processing 500,000+ daily orders. Marketing spend was significant and growing, but the company lacked the ability to evaluate campaigns on anything beyond acquisition cost. There was no visibility into whether the users a campaign brought in were actually worth what was paid to acquire them.

Build a system that could connect acquisition data to long-run user behaviour, enabling the marketing team to evaluate campaigns not just on CAC but on whether the lifetime value of acquired users ever exceeded what was spent to acquire them.

The attribution problem at Blinkit was more complex than a standard LTV model because ad spend fell into two distinct categories: acquisition campaigns targeting new users (CAC) and retention campaigns targeting existing users to drive repeat orders (CRC). A sale could not simply be attributed to the campaign that acquired the user — users were continuously exposed to ads throughout their lifecycle, and each individual sale needed to be attributed to the last campaign the user had engaged with in the seven days prior to that order.

This meant the total advertising cost attributable to a user accumulated over time as both CAC and ongoing CRC. The measure of profitability was whether cumulative lifetime revenue exceeded cumulative (CAC + CRC) — a substantially harder bar to clear than CAC alone. The data foundation was raw ad engagement data from AppsFlyer, joined to the order database using a last-touch seven-day attribution window and rebuilt into cohort-level models tracking running cost-of-revenue over time.