How position based attribution works
Position based splits a conversion by where each touch sits in the journey rather than by when it happened. The first touch and the last touch each take the largest share, and everything in between divides what is left. Because the weighting is heavy at both ends and light in the middle, it is usually drawn as a U and often called U-shaped attribution.
The reasoning is that two moments in a journey are structurally different from the rest. One introduced the person to you and one closed the deal; the visits between are maintenance. Whether that is true of your customers is an assumption you are making, not something the model has discovered.
Some tools let you change the weighting, others fix it. Either way the split is arbitrary in the strict sense: it was chosen because it seemed sensible, not derived from your data. If a journey contains only one touch, that touch takes everything, and with two touches the middle share has nowhere to go and is absorbed by the ends.
Why position based attribution matters
It is the compromise most people can actually agree on. A final-click rule tells the closing team they did everything; a first-click rule tells the awareness team the same. Position based pays both, which makes budget conversations less circular and stops a discovery channel being cut simply because it never appears at the end.
It also travels well between businesses. It needs no training data, works at low conversion volumes, and gives the same answer every time from the same paths, which matters when you have to explain a number to a client months later.
Common mistakes with position based attribution
The commonest is presenting the weighting as a finding. Nothing in your account proves the middle touches deserve the smaller share; that was decided before the data was read. If a middle-of-journey channel is the one answering the questions that make people buy, this model will systematically underpay it.
The other is applying it to journeys that were never U-shaped. Repeat purchases, renewals and long research cycles often have no single decisive entrance, and forcing a first-and-last frame onto them creates confident numbers about a shape that is not there.
How to act on it
Look at your path reports before choosing it. If most journeys are short, this model collapses into a final-click view and the extra complexity buys you nothing. If most are long and varied, check whether the middle is genuinely quiet before you pay it least.
Use it as a shared language for budgeting rather than as a bidding signal, and hold the middle channels to a separate test — a period with the channel off tells you more than any weighting. If your paths look shorter than the journeys your customers describe, fix the measurement first through a proper analytics and tracking setup, because every shared model gets worse as paths get truncated. Where you have the conversion volume for it, data-driven attribution derives the weighting from your own account instead of assuming it.
