Ask a marketing tool where a customer came from and it will give you one answer, delivered with total confidence. The problem is that the honest answer is a journey: a directory listing read in September, a comparison page skimmed twice, an ad clicked in October. Compress that into a single stamp carelessly and you will defund the channel that started the journey to feed the one that ended it.
Swipe Again keeps the money stamp simple and permanent: acquisition is credited last-click, forever, because payouts and budgets need one stable rule. But the stamp is not the whole record. Alongside it we store the first identified touch and the full touch trail for every merchant — classified from ad click IDs, referral parameters, and UTMs, kept for ninety days in a first-party cookie — so the journey survives even though the stamp stays singular.
The interesting event is the mismatch. When a merchant's first touch is a different channel than the one that got stamped, that is an assist: the listing that warmed them up for the ad that closed them. Our system flags it once per merchant so it gets noticed rather than averaged away. In reporting, the same data can be re-sliced — first-click view, or a position-based split that gives the opening and closing touches forty percent each and divides the rest among the middle — while business totals stay identical under every model. The views are for learning. The stamp is for money. Confusing the two is how attribution systems start editing history.
Small teams cannot afford data-driven attribution; it needs conversion volume they do not have. What they can afford is refusing to throw the journey away. Capture every touch now, decide with models later. Trails you did not store can never be backfilled — a lesson the recovery flow applies to failed payments too, where ninety days of history is the difference between a guess and a schedule.
Top comments (0)