Free attribution model tool

An attribution model is a rule for splitting the credit for a conversion across the marketing touchpoints that came before it. Someone might find you through organic search, return from an email, click a paid ad and then buy. That is one sale and three channels, and the model decides who gets what share. Different models give different answers to the same data, because there is no single correct split, only the rule you chose.

Last-click attribution gives all the credit to the final touchpoint before the conversion, and first-click gives it all to the first one. Last click favours channels people use when they already intend to buy, such as brand search and retargeting. First click favours discovery channels such as organic search, social and display. Both are wrong in the same way: they hand 100% of the credit to one interaction in a journey that had several.

Multi-touch attribution spreads credit for a conversion across every touchpoint in the journey rather than giving it all to one. Linear splits it evenly, time decay weights touchpoints nearer the conversion more heavily, and position-based typically gives 40% each to the first and last touch with the rest shared between the middle. It gives a fairer picture of how channels work together, but the weightings are still assumptions you picked rather than anything measured.

Data-driven attribution works out the credit split from your own conversion data instead of applying a fixed rule. Algorithms such as Markov chains measure how much your conversion rate drops when a channel is removed from the journey, while Shapley values calculate each channel's average contribution across every possible ordering. The result reflects how your customers actually behave rather than a rule someone chose. It needs a decent volume of conversions to be stable, so low-traffic accounts get noisy answers.

No model is accurate in the strict sense, because attribution is a credit allocation, not a measurement. Data-driven models such as Markov and Shapley come closest to reflecting real behaviour, since they are derived from your journeys rather than assumed. That said, they can only see touchpoints you track, so offline exposure, word of mouth and view-through effects are invisible to all of them. The usual practice is to compare several models and act on where they disagree.

Last non-direct click gives the credit to the last touchpoint before the conversion that was not direct traffic. It exists because direct usually means someone typing your URL after another channel introduced them, so crediting direct tells you nothing actionable. It is the default in many analytics tools for that reason. It is still last-click logic with one exception bolted on.

Because the model decides which channels look like they are working, and budgets follow that. Last click systematically underfunds awareness channels, which rarely appear at the end of a journey, and overfunds brand search and retargeting, which almost always do. Switch models and the same spend can look like a success or a waste. Comparing models before reallocating budget is the point of doing attribution at all.

Thirty days is the common default, and it suits most ecommerce. Considered purchases with long research phases, such as B2B software or cars, often need 60 or 90 days to capture the first touch at all. A window that is too short makes discovery channels disappear, while one that is too long credits touchpoints that had nothing to do with the sale. Set it from your actual time-to-purchase rather than the default.