How linear attribution works
Linear takes a single conversion and divides it evenly across every recorded touchpoint inside the lookback window. If someone met you through four touches before enquiring, each of those four keeps an identical quarter of the sale and an identical quarter of its value. Nothing about the order, the timing or the kind of touch changes the split.
That makes it the least opinionated of the shared models. Where a data-driven model tries to work out which touches actually moved the person along, linear refuses to guess and treats presence in the journey as the only thing that counts. It needs no training data and produces a stable answer from any volume of conversions, which is why it is often the first shared model a small account can use honestly.
One consequence catches people out: because credit is fractional, channel reports stop showing whole conversions. A channel can show a part of a conversion, and the parts across channels add up to the one real sale. That is the model working correctly, not a rounding error.
Why linear attribution matters
It makes assisting channels visible without any modelling you have to defend. Email, retargeting, organic articles and social posts that never close a sale but appear constantly on the way to one finally register as contributing something, and the argument moves from whether they matter to how much.
It is also easy to explain to a client or a board. Everyone in the journey gets an equal share; there is no black box to trust and no weighting anyone can accuse you of choosing to flatter a channel you happen to run.
Common mistakes with linear attribution
The main one is forgetting that fairness is an assumption, not a finding. A cheap, high-frequency touch that appears in nearly every journey collects credit simply by being everywhere. Retargeting and branded search are the usual beneficiaries: they turn up late, after the decision was effectively made, and linear pays them the same as the ad that created the interest.
The second is bidding on it. Ad platforms need feedback tied to the touch they can influence; feeding them equal fractions across a long path makes the signal vague and the learning slow.
How to act on it
Use linear as a discussion model rather than a decision model. Run it alongside a final-click view over the same period and look at which channels gain the most credit when you move from one to the other — those are the assisting channels, and that gap is the real finding.
Before trusting the split, make sure the journeys behind it are complete. Missing UTM tags, untracked email links and lost cross-device visits all shorten paths, and a shorter path hands each remaining touch a bigger share of a sale it may not have earned. Getting the tracking foundations right matters more here than the choice of model, and where a channel still looks suspiciously good, test it by pausing it rather than by arguing about weightings. The alternative worth comparing is time decay, which weights touches nearer the conversion.