What Data-Driven Attribution Is
Data-driven attribution is an attribution model that learns how to share credit from your own data. Rule-based models fix the split in advance: last touch gives everything to the final click, linear splits evenly, time decay favors recent touches. A data-driven model looks at many real journeys, the ones that converted and the ones that did not, and works out how much each channel actually moves the needle.
The result is still multi-touch attribution: credit is spread across touchpoints. The difference is that the split comes from evidence rather than a formula someone picked.
How the Main Methods Work
Markov chain models treat each journey as a path through channels, ending in a conversion or a drop-off. From thousands of paths, the model learns the probability of moving from one channel to the next and the overall chance of converting. It then removes one channel at a time and recalculates. If removing paid social cuts modeled conversions by 30%, paid social has a large removal effect and earns a large share of credit. Credit is the removal effects scaled to add up to total conversions.
Shapley value models come from cooperative game theory. They consider every combination of channels, measure the conversion rate with and without each channel, and give each channel its average marginal contribution. They are fair in a formal sense but usually ignore the order of touches.
Platform models are what ad networks run in their own dashboards. Google made data-driven attribution the default in Google Ads and removed most rule-based models from Google Ads and Google Analytics 4 in 2023. These models only see the platform's own touchpoints, so they describe journeys inside one network rather than across your media plan.
What It Is Good For
- Budget mix. Data-driven credit shows which channels assist conversions they rarely close, such as video or creators that lead to brand search.
- Spotting channel synergy. Paths where two channels together convert far better than either alone suggest the channels should be planned together.
- Fairer channel comparisons. Retargeting and brand search often look outstanding under last touch because they sit at the end of journeys. A data-driven view tempers that.
Its Limits
Data-driven attribution is still correlational. It measures how conversions change in the data when a channel is absent from paths, not what would happen if you actually stopped spending. Channels that reach people who were going to convert anyway still earn credit. For causal answers, use incrementality testing with a holdout group.
It also depends on seeing the journey. Touches lost to privacy rules, consent choices, device switches or SKAN aggregation do not appear in paths, and the model cannot credit what it cannot see. On iOS in particular, a large share of installs may never have a user-level path.
Finally, it is harder to reconcile. Networks bill on their own models, so data-driven credit will not match any network dashboard. Use it for planning, and keep last touch for daily operations and postbacks.
Data-Driven Attribution in Linkrunner
Linkrunner credits each install to one source with last-touch rules, so postbacks and network reconciliation stay consistent. On top of that, multi-touch attribution is available per project on request. A data-driven model based on the Markov removal effect runs daily per project, and you can switch between last click, platform-reported, multi-touch and a hybrid model. Reports show assists, journeys and channel synergy, and cross-device journeys connect through your customer user ID and hashed email or phone, never IP. The default lookback is 30 days, with a 7-day revenue window. See analytics and ROAS measurement.
Getting Value from It
- Compare data-driven credit with last touch channel by channel, and focus on the biggest differences.
- Refresh on a schedule and look at trends rather than one day's numbers.
- Group low-volume channels so their credit is stable.
- Confirm big budget moves with an incrementality test before committing.