a real campaign, start to finish: what numu saw that the team couldn't, what it changed, what it left alone — and the number at the end.
a studio came to us with a live RPG spending mid-six-figures a month on meta across six markets. CPI in thailand and the philippines had drifted 22% north of target, and the team couldn't tell whether it was creative fatigue or genuine market pressure. same question every UA lead is stuck on at 11pm. here's the whole six weeks.
week 1 — what numu saw
on the first full read, numu flagged creative fatigue on the top-spending video in TH: frequency 4.8, CTR down 48% over 72 hours, while source quality was still above benchmark. translation: not a market shift, just a tired creative — the exact thing a human analyst would have caught a week and several thousand dollars later. the PH softness traced to the same video, cloned into that market.
weeks 2–3 — what changed
numu paused the fatigued creative, promoted two under-served variants that were quietly beating it on hook rate, and shifted $400/day out of TH and back into indonesia, where CPI was $0.47 and there was real headroom. in parallel it briefed and shipped two TH-localized variants — grounded in the game's own assets — inside four days.
what it deliberately left alone
the bid strategy on the two markets that were already at target. numu is conservative by design: it doesn't touch what's working just to look busy, and anything that affects brand goes through the critic layer first. no theatrics, no dashboard full of blinking changes for their own sake.
the number
| metric | start | week 6 |
|---|---|---|
| blended CPI | 22% over target | 30% below start |
| D7 ROAS | baseline | +61% |
| UA lead's week in dashboards | most of it | near zero |
blended CPI down 30% in six weeks, D7 ROAS up 61%. the part the studio actually noticed: their UA lead spent those six weeks on partnership deals instead of babysitting ad sets. want the same read on your account?
