Subscription meal kit, Bangkok metropolitan area
A six-year-old meal kit had a “loyal” cohort that looked stable on a rolling 90-day repeat chart. During Predictive CLV for Commerce Teams, the team rebuilt the same buyers by first-plan type and by whether the first box included a free protein add-on.
The free add-on cohort returned at a similar rate but at a contribution that never recovered shipping. The valuation note ranked that segment last, against internal habit. Media paused lookalikes built from that group. Three months later, paid CAC on the remaining plans did not magically fall; it simply stopped being compared to a fictional lifetime. The team still has not automated the sheet. They open it every planning cycle.