Case #052: The scale that didn't blow up

Case #052: The scale that didn't blow up

A sweepstakes campaign sat at $90/day, 80% ROI, for six days. Stable. Boring. Begging to be scaled. Every buyer knows the next part: you 5x the budget and the ROI evaporates overnight. I've torched four winners that way. So I built a scaling ladder and never skip a rung.

The ladder, one step per 48 hours:

— Step 1: raise budget 20%, not more. The auction re-learns; bigger jumps reset the algorithm's confidence and you pay tuition again.

— Step 2: before each raise, confirm yesterday's ROI held within 10 points of baseline. Drifted? Hold the rung, don't climb.

— Step 3: when one ad set caps out, duplicate it into a fresh set rather than pushing the original harder. Spreads the auction load.

— Step 4: every third rung, refresh one creative. Frequency above 2.5 on the same audience is the silent ROI killer.

— Step 5: keep a 'pull-back number' written down — the spend level where ROI last dipped below 50%. If you hit it, drop one full rung, no ego.

Day 1: $90/day. Day 14: $310/day, and ROI only slipped from 80% to 61% — a fee I'll pay all day. Total over the run: $2,980 spent, $4,790 back, 60% ROI. The version of me from two years ago would have jumped to $450 on day 2 and watched it die by day 4.

The lesson: scaling isn't a bigger bet, it's the same bet made slowly enough that the algorithm never panics.