Multi-Touch Attribution

Multi-touch attribution (MTA) divides credit for a conversion among the ads and channels a person interacted with on the way to buying. In Google Analytics 4, this shows up in the Advertising section, under Attribution. Google's help page lists three reporting models: data-driven attribution, paid and organic last click, and Google paid channels last click. The Attribution paths area includes a report on key event paths, which shows common sequences of touchpoints leading to a key event. Other platforms offer similar path views, with their own models and settings. In GA4, you can switch the reporting attribution model, and Google says the change applies to historical and future data.

The method works at the level of an individual path. A system records touchpoints for a user, such as a Meta ad view, an email click, and a Google search, then applies a rule or a model to split credit. Rule-based models give fixed shares, for example all credit to the last click. Data-driven models use machine learning to estimate how each touchpoint changes the likelihood of conversion, using both converting and non-converting paths, and they give fractional credit that sums to 1. Google Analytics removed the first click, linear, time decay, and position-based models in November 2023, leaving data-driven attribution and two last-click models. The key event paths report shows common sequences of touchpoints, which is useful for seeing how people move between channels before they act.

MTA interacts with what the system can see. It needs touchpoints linked to the same person across channels and devices, and each link depends on identifiers, consent, and platforms sharing data. Where that fails, the path is incomplete. Walled-garden platforms report their own views, clicks, and conversions, and they often claim the same sale. Ad blockers, browser limits on cross-site identifiers, and app tracking choices remove touchpoints from the path. The result is credit concentrated on channels that still report cleanly, usually the last measurable click, such as paid search, and less on upper-funnel channels that influenced the decision earlier. Compare data-driven with last click for each channel. A channel that gains under data-driven may be doing earlier work than last click shows.

The main mistake is treating attribution as proof of cause. A model that gives 30 percent credit to a social ad shows that the ad appeared in the path, not that the sale would have been lost without it. Another mistake is comparing numbers across tools that use different windows, models, and definitions, and expecting them to match. A third is changing the model and concluding that performance changed, when only the credit moved. A fourth is ignoring unobserved paths: offline visits, word of mouth, and untracked views are not in the data, so the paths are partial. Keep in mind that offline conversations, word of mouth, and untracked views never appear in the path, so the data is only a partial picture.

Use MTA for what it does well: comparing campaigns and keywords within the channels you can track, and seeing common path patterns. In GA4, open the model comparison and key event paths reports and look at how credit shifts between data-driven and last click. Channels that gain under data-driven attribution may be doing earlier work. Then test the largest claims with an incrementality test or a geo test, and use aggregate checks such as total revenue against total spend. Treat MTA as a lens on tracked behavior, not a measure of cause. Document the model, window, and conversion definition in every report so numbers from different weeks can be compared fairly.