Data-Driven Attribution

Data-driven attribution, often shortened to DDA, is a way of splitting credit for a conversion across the ad interactions that came before it. In Google Ads it is a setting on each conversion action, and Google's help page says it is the default for most conversion actions. In Google Analytics 4 it is one of three reporting attribution models. Rather than following a fixed rule such as 'last click wins', it uses your account's past data for that conversion action to estimate how much each interaction contributed, and gives each one a share. A path with three ad clicks might split credit 50, 30, and 20 percent, with the exact split coming from the model. You can find the setting on a conversion action's page in Google Ads, and in GA4 under the attribution settings for your property.

Under the hood, the model compares what happened on paths that ended in a conversion with what happened on paths that did not. If people who saw a certain keyword or ad early on convert more often than similar people who did not, that interaction earns more credit. Credit is fractional, so a single conversion can show up as 0.4 of a conversion for one keyword and 0.6 for another, which is why conversion columns sometimes show decimals. In GA4 the help page notes that conversions can be reattributed for up to 7 days after the conversion. The Google help pages checked here state no minimum conversion count, so treat any exact threshold you see elsewhere as unconfirmed. Because credit is fractional, reports may show 12.4 conversions rather than 12. That is the model sharing one sale across several interactions, not a counting error.

DDA matters because of what it feeds. When you use Smart Bidding, the conversion credit assigned by your attribution model is what the bidding system learns from, so under last click it learns to value final-click keywords and under DDA it learns to value earlier touches too. That can shift budget toward generic or upper-funnel keywords that never get the last click. DDA is also confined to the platform measuring it. Google Ads credit covers Google Ads interactions, and GA4 data-driven credit uses GA4 data, so neither sees a leaflet, a podcast, or a Meta ad. If you are moving from last click, expect branded and retargeting keywords to look slightly weaker and generic or discovery keywords to look stronger, though the size of the shift depends on your own paths.

The usual mistakes are treating DDA as the truth and treating it as a black box you cannot question. It is a model, and models can be wrong, especially with thin data. Another mistake is comparing last-click and data-driven numbers from two dates and concluding performance changed, when only the credit rule changed. Switching the model also changes how bidding sees past conversions, so expect a change in reported keyword performance. Many people also expect DDA to find a hidden channel; it can only redistribute credit among interactions it can observe. Remember that the model learns from the past. A new campaign or channel with little history can be undervalued until it has built up data, so give it time before judging. Do not forget conversion lag.

To use it, open the attribution model comparison in Google Ads or the model comparison report in GA4, and view the same conversions under last click and data-driven side by side. Keywords whose credit rises under DDA are the ones that help earlier in the path. Keywords whose credit falls were mostly collecting the last click. This week, pick your five biggest keywords or campaigns, compare their conversions under each model, and flag any you were about to cut that look better under data-driven. Then validate big budget moves with a holdout test, not the model alone. Pair the model with a simple sanity check: total conversions in your store or CRM should still roughly match the sum across channels, since attribution moves credit but should not create sales.