Case #011: The day I stopped reading days

Case #011: The day I stopped reading days

A mobile game install offer with an event-based payout: $0.40 on install, plus $6 if the user reached level 10 within 7 days. $2,800 budget. The trap is that daily ROI is meaningless here — money spent today pays back across the next week, so any single day's dashboard is comparing today's spend against last week's maturing revenue.

I nearly killed it on a normal Tuesday. The day-view showed $260 spent, $190 back, red. But I'd been forced to think in cohorts by the payout structure, so instead of reading the calendar I read the install dates.

The Monday cohort told the real story: 1,100 installs at $0.44 each, $484 spent, and by day 7 those same installs had thrown off 47 level-10 events at $6, plus the install payouts — $766 back from one day's buying. The campaign was profitable; the dashboard just couldn't show it because spend and revenue lived on different clocks.

Once I tracked by cohort, I could see which days of traffic aged well and which decayed, and I cut the two weekday sources whose users never reached level 10.

Full arc, measured at full cohort maturity, 19 days of buying: $2,800 spent, $4,510 back. 61% ROI — invisible on any daily view, obvious the moment I lined spend up with its own cohort.

The lesson: when revenue matures over days, a daily dashboard isn't conservative, it's wrong — read cohorts or you'll kill your winners on a slow Tuesday.