Budgeting AI Generation Like Media Spend: The ZAP Model

Per-seat SaaS pricing made sense when software was a tool you used. Generative AI is different: every image, video and copy batch has a real compute cost, and usage varies 100× between a quiet week and a launch week.

Credit systems — like Boosta’s ZAP — price the work, not the seat. You pay for what you generate, top up when launches demand it, and idle months cost you a subscription floor instead of an inflated per-user bill.

An image generation costs 25 ZAP; a video costs 100. But the number that matters is upstream: how many generations does it take to produce one asset that beats your incumbent in a fair test?

If one in six images becomes a winner, your cost per winning static is roughly 150 ZAP — a few pounds. Compare that to a £400 agency asset with the same one-in-six hit rate and the economics are not close.

Allocate roughly 5–10% of monthly media spend to creative generation and testing. Teams spending £10k/month on ads should comfortably invest the equivalent of £500–£1,000 in producing and testing new concepts — the leverage on the other 90% is enormous.

Track credit burn by campaign, not by month. A launch that consumes 3,000 ZAP and finds two scalable winners is cheap. A quiet month that burns 500 ZAP on aimless generation is expensive.