The brief
The client was running two subscription properties in Japan with steady traffic and steady offers, but inconsistent creative performance. They’d commission a batch of 40 banners every six weeks from a freelance pool, the batch would perform well in week one, then decay sharply. By week four they were running creative no one wanted to click on, which was poisoning their auction position.
The ask was simple: stop the cycle.
The approach
The root cause was batch-and-burn. Creative shipped in monthly chunks meant every banner in the batch shared visual vocabulary, copy patterns, and pacing. Users were getting tired of the style, not just the individual ads.
We split the 40-banner monthly volume into weekly drops of 8–10 banners, each drop pulling from a different angle in a six-angle rotation. Every week looked visually distinct from the previous one. We also built in deliberate variance within each drop — same offer, three different copy registers — so the auction algorithms had material to pick winners from without us guessing.
The strategy review at the end of month one swapped two angles that weren’t pulling for two new ones.
The result
CTR stabilized above 2.1% within the first three weeks and has held there for 12 consecutive weeks. The client’s media buyer told us the auction position metric (impression share at target bid) recovered to where it had been in mid-2025 within a month.
They renewed for a 6-month commitment after the original 3-month minimum.