What our scan layer actually sees.

We scan ads. A lot of them. Right now: 47 customer brands plus their named-and-suspected competitors, ~12,400 unique ad creatives indexed across Meta Ad Library and TikTok per day, refreshed nightly. The math: a typical DTC niche has 8-15 active brands running ads, each with a rotation of 30-80 live creatives. Multiply through and you're looking at thousands of hooks across any single category.

Of those 12.4k daily, ~1.8k are net-new vs the prior week. That's the signal. New creatives = either the brand testing fresh ideas, or scaling something that started winning. We weight new entries by reach + engagement velocity over their first 48 hours; the ones that climb fast are the ones that triggered something.

Our decoder looks at each rising creative across five dimensions: hook archetype (POV / GRWM / problem-reveal / unboxing / etc.), scene type (mirror selfie / propped medium / walking / static), voice cadence (slow-confident / fast-conversational / deadpan / hyped), audience trigger (the inferred who-is-this-for), and pacing (where the dwell happens, how long before the brand mention).

Those five dimensions are what we call the decoded pattern. When three competitors in your category all hit on the same combination — say, POV self-narration + mirror selfie + 195 wpm conversational + audience-trigger="new mom" — that's not a coincidence. That's a winning pattern in your category. We surface it. Your brief inherits it. Your variants ship under it.

What we don't do: we don't sell the raw scan as a standalone product. The reason is that raw competitor monitoring is a commodity — the value isn't in seeing the data, it's in turning the data into clips on your brand. Our scan layer exists to feed our generation layer. Buying just the scan would be like buying just the engine.