Myth: a longer attribution window gives a more complete, more accurate picture

Myth: a longer attribution window gives a more complete, more accurate picture

The question: extending your conversion window from 7 to 30 to 90 days captures more of the journey — so longer must mean more accurate, right?

What the trade-off actually is: lengthening the window doesn't monotonically increase accuracy; it trades one error for another. Short windows truncate genuinely-influenced late conversions (under-credit). Long windows sweep in conversions that would have happened anyway and attach them to an old, possibly irrelevant touch (over-credit, often badly). You are choosing between two biases, not climbing toward truth.

The nuance: long windows amplify the brand/retargeting over-crediting problem, because the longer you look, the more likely a high-intent user passed through some cheap late touch you'll now reward. They also degrade under data decay — cookies expire, sessions get mis-stitched, and the oldest touches in a 90-day path are the least reliable. And the 'right' window is not universal: a SaaS trial cycle, an impulse e-commerce buy, and a car purchase have wildly different true consideration lengths, so a single window misfits most products.

Bottom line for practitioners: pick the window from your actual time-to-conversion distribution (look at the empirical cumulative curve and cut where it flattens), not from a desire for 'completeness.' Then recognize that no window resolves the credit problem — it only shifts the bias. The completeness you actually want comes from incrementality testing, which doesn't care how long the observed path is.