30 creative variants × thousands of campaigns

Context

An adtech platform generates 30 creative variants for each of thousands of campaigns. The per-variant cost is tiny, but multiplied across campaign volume it becomes the number that gates the feature.

The decision

Economics before architecture: unit cost × volume decides if the variant feature is viable — model choice, caching, and batching are the levers that gate the roadmap.

What most miss

Engineers design the variant pipeline before anyone models the unit economics; at this volume the cost decides the feature, not the other way around.

Stakes

Ship the feature without the unit-economics model and a tiny per-call cost becomes a budget surprise at campaign scale.