01Insights  /  Insight

Attribution is not one model, and never was

Every attribution method is wrong in a specific, predictable way. Knowing which way is the whole job.

Deterministic

Precise where it can see, blind everywhere else. Good for the last step, poor for the screens that created demand.

Probabilistic

Fills the gaps, at the cost of certainty. Useful for direction, dangerous when quoted to two decimal places.

Test and learn

The closest thing to proof available, and the most often skipped because it takes planning.

Using them together

Each method is a different instrument reading the same system. Agreement between them is the signal worth acting on.

WireframeSite map