Attribution windows are among the least examined settings in advertising measurement and among the most consequential. Changing one alters which channels appear to be working without anything about the campaigns having changed.
The window defines eligibility, not causation
A window states how long after an interaction a conversion may still be credited to it. Anything falling outside is attributed elsewhere or not at all.
The setting makes no claim that the interaction caused the conversion. It only decides which records are permitted to be connected.
Because the choice is arbitrary within a reasonable range, two teams measuring identical activity with different windows reach different conclusions honestly.
Length favours different channels systematically
Short windows favour channels that appear late in the decision, such as branded search and retargeting, where the conversion follows quickly.
Long windows give credit to channels that create awareness well before purchase, which otherwise show almost nothing at all.
Neither setting is neutral, and choosing one without recognising the bias amounts to deciding the answer before the measurement begins.
View-based windows are the contested territory
Crediting a conversion to an advertisement that was seen but not clicked assumes the exposure contributed, which is plausible and difficult to verify.
Long view windows can credit almost any conversion to almost any display campaign, since the audience overlap between browsing and buying is large.
Advertisers who shorten these windows usually see reported display performance fall sharply, which reflects the accounting rule rather than a change in effectiveness.
The purchase cycle should set the length
A window shorter than the typical consideration period will systematically undercount, and one much longer will absorb conversions that would have happened regardless.
Measuring the actual distribution of time between first interaction and purchase gives a defensible basis for the setting.
That distribution differs between products within the same business, which is why a single account-wide window rarely fits everything it is applied to.
Consistency matters more than correctness
Since no window is objectively right, the practical requirement is that the same one is used across the comparisons being made.
Changing a window mid-period produces a step change in reported performance that is easily mistaken for a campaign effect.
Recording the setting alongside the numbers, and re-baselining deliberately when it changes, prevents an accounting adjustment from being read as a result.