Myth: linear attribution is the 'fair' model because it treats every touch equally

Myth: linear attribution is the 'fair' model because it treats every touch equally

The question: linear attribution splits credit evenly across all touchpoints. Equal sounds unbiased — so isn't it the neutral, fair default?

What the reasoning shows: 'equal' and 'fair' are not the same word. Linear attribution makes a strong and almost certainly false assumption — that a fleeting display impression and a 10-minute comparison-page session contributed identically. By forcing equality, it doesn't remove bias; it imposes a specific, wrong prior that every touch matters the same regardless of channel, timing, or engagement depth.

The nuance: linear's flatness has a predictable distortion. Because it credits every touch equally, it mechanically rewards channels that generate high touch volume — frequent retargeting, repeated email sends, low-cost impressions — over channels that appear once but do heavy persuasive lifting. A single decisive review-site visit gets the same 1/N as the eighth retargeting banner. So 'fair' linear quietly subsidizes spammy, high-frequency tactics.

There is no neutral attribution model. Every model encodes assumptions; linear's assumption (uniform contribution) is just more hidden than last-click's.

Bottom line for practitioners: stop reaching for linear as the 'safe middle.' If you want a heuristic that hedges position effects, a U-shaped curve is at least defensible. But recognize that any credit-split model encodes a theory of how persuasion works — pick the one whose assumptions you can defend, and validate channel value with experiments rather than appealing to false neutrality.