30-day repeat order rate had plateaued at 40% — well short of where a maturing food-delivery market should sit. I built a CY2026 retention strategy anchored on one OKR: take repeat orders from 40% to 62%, grounded in a 7-stage funnel teardown and 4 paired user cohorts.
A flat repeat-order curve usually means one of two things: the product isn't worth returning to, or something specific in the post-order journey is quietly losing people. I needed to know which, before writing a single initiative.
Rather than treating "retention" as one number, I broke the journey from discovery to a second order into 7 discrete stages — so any fix could be pointed at the exact place intent breaks down.
A single retention curve hides more than it reveals. I paired active-vs-lapsed and loyal-vs-price-sensitive users across the same 7 stages, so the plan would target what actually predicts a second order — not what merely correlates with one.
Baseline for "what good looks like" at each of the 7 stages — the behaviors the other cohorts are missing.
Paired against Cohort A to isolate exactly where the journey diverges, stage by stage.
Tests whether rewards mechanics alone explain repeat behavior, or whether journey friction still dominates.
Paired against Cohort C to separate "retained by habit" from "retained by discount."
40% repeat order rate
One-size-fits-all retention emails, sent on the same schedule to every user regardless of cohort or funnel stage.
62% repeat order rate
Cohort-personalized nudges, each targeted at the specific funnel stage where that cohort actually drops off.