Attribution

The system links a conversion to earlier touches within a lookback window and then applies a credit rule. The rule may give all credit to one touch or share it across many.

Run two models against the same data and compare the credit shifts. The differences show how much the story depends on the rule and not on customer behavior.

A running-shoe retailer logs one order of $150. The buyer clicked a paid search ad on Monday, read a blog post from an organic search on Wednesday, and opened an email on Friday before buying.

Last-click gives the email $150. First-click gives paid search $150. A linear model gives each about $50. None of these is wrong; they answer different questions, and the retailer pairs two of them in its monthly review.

Attribution is the umbrella idea of assigning credit. Last-click attribution is one specific rule that gives it all to the final touch. The umbrella has many models.

Treating a model as truth. Switching models after seeing results. Ignoring touches that are not tracked. Using too short a lookback. Assuming credit equals cause.

Choose one model as your standard and one as a cross-check. Report both. When they disagree sharply, investigate the channel that moves.