Anatomy of a Verdict: the Coupon That Beat Its Baseline
The companion receipt: a marketplace coupon that beat its pre-period baseline by a wide margin and held the lift after the promotion ended — a closed loop graded a hit, from the same ledger that logs the misses.
Published August 18, 2026
This is the companion to the campaign we killed, and it is deliberately the opposite outcome. A marketplace coupon on a DTC brand we operate was proposed with a prediction, ran its window, beat its pre-period baseline by a wide margin, and held a meaningful share of that lift after the discount ended. Graded a hit, kept in the ledger. Same machine as the miss — that is exactly why it counts.
Birth: a baseline, chosen in advance
Before the promotion ran, the action recorded what it was measuring against: the product's own trailing sales baseline, and an explicit expectation that a lift worth having would outlast the window. Choosing the baseline before seeing the result is the whole trick — it is measurement at birth — because a baseline picked after the fact can always be drawn to flatter the outcome.
Life: the lift, and the tell
During the window the coupon beat its baseline comfortably. That alone proves little; a discount lifting sales is not news. The tell came after: when the price returned to normal, a real portion of the elevated demand stayed. New buyers had found the product, and some of them stuck. The during-number was the headline; the after-number was the evidence.
Death: a hit, recorded next to the misses
On its due date the loop closed with a positive verdict, filed in the same outcome ledger that holds the search campaign's miss. Both are dated, both are checkable, neither can be quietly edited later. The credibility of this win rests entirely on the visible presence of that loss two shelves over.
Why we publish the pair
A single receipt is an anecdote. A win and a miss from the same week, scored by the same rules, is a method — and a method is the thing a buyer can actually trust. This is what the eleventh question looks like answered: not “our agents are impressive,” but “here is one bet that paid and one that did not, and here is precisely how we knew.”
Questions founders ask
- How do you prove a promotion actually worked?
- You compare the result to a baseline you defined before running it. A coupon that "did well" means nothing without the pre-period number it beat. Here the marketplace promotion was measured against its own trailing baseline; it beat that baseline by a wide margin during the window and — the part that matters — held a meaningful share of the lift after the discount ended, which is the signal that it pulled in demand rather than just pulling sales forward.
- Why does the after-the-promotion number matter more than the during?
- Because almost any discount lifts sales while it runs; that is not evidence, it is arithmetic. The question that separates a good promotion from an expensive one is what happens when the price goes back up. A lift that persists past the window suggests new customers and durable demand. A lift that vanishes the moment the coupon ends was a subsidy on sales you would likely have made anyway.
- Is this the same machine that killed the search campaign?
- Yes — same seats, same gate, same measurement-at-birth discipline. One bet was graded a miss and shut off; this one was graded a hit and kept. That is the point of showing both: an outcome ledger is only trustworthy because it lands on both sides of the line. A system that only ever reports wins is a marketing surface, not a ledger.
- Are these exact figures?
- No. The event is real and drawn from our ledger, but the numbers are rounded and directional and no brand, product, marketplace account, or person is named, per our client-data rule. The discipline is public; the ledger's contents are private and verified in a data room, not here.