Trade spend. The promotions that paid, not the ones that ran.
Trade spend is the second-largest line on the P&L and the one nobody can defend. Next year’s calendar is mostly last year’s calendar: whatever ran gets run again, because sales went up during the window and there is no baseline to say what would have happened anyway.
Eridian builds the baseline first — what each SKU would have sold without the promotion — then measures true incremental lift, net of forward-buy and cannibalization. The same Ontology validates deductions against the promotion that earned them, so planning and settlement finally run on one set of numbers.
A counterfactual baseline per SKU and account, so lift is measured against what would have happened — not against the four weeks before.
True incremental lift by promotion, net of forward-buy and cannibalization, with the promotions that lose money named.
Deduction validation against the promotion that earned it — every one, not a sample — so the trade line reconciles instead of surprising you.
The objects this use case reads and writes — stood up during the diagnostic, shared with every use case that follows.
This is a diagnostic candidate: two to three days on your floor, the relevant slice of the Ontology stood up from your data, and a working first pass you can judge in production terms — not a deck.
AI stopped
being a demo.
Nobody sent a memo.
Not sure where to start? Book a half-hour with an engineer who ships frontier AI weekly. You’ll leave knowing what’s possible and what’s still a demo.
with an engineer