Attribution · Google Ads
Data-driven attribution is now the default model in Google Ads, and every other model except last click has been retired. It is a genuine improvement over last click. It is also confined to Google's own touchpoints, which means it solves a smaller problem than most advertisers think it solves. Here is exactly what DDA does, what the docs require, and where the real gap remains.
Under last click, the final ad someone clicked takes 100% of the credit and everything before it takes none. Data-driven attribution replaces that with something considerably smarter: it compares the paths of customers who converted against the paths of customers who did not, finds the interactions that reliably show up ahead of conversions, and redistributes credit toward them. Each model is specific to your account and built from your own conversion data, so two advertisers in the same industry get different models. Google's own example is a tour operator who discovers that people who click a broad "Bike tour New York" ad before a specific "Bike tour Brooklyn waterfront" ad convert at a higher rate, so credit shifts back toward the broad term that last click was giving nothing to. That is a real correction, and it is the reason DDA usually surfaces upper-funnel keywords that looked worthless.
First click, linear, time decay, and position-based are no longer supported. Conversion actions that used them were upgraded to data-driven. You have two choices now: data-driven or last click. That is the whole menu.
Clicks and video engagements on Search including Shopping, YouTube, Display, and Demand Gen, across website, store visit, and Google Analytics conversions. Anything outside that set is not part of the model.
Every conversion action is eligible regardless of volume, but Google recommends at least 200 conversions and 2,000 ad interactions within a 30-day period for the model to identify patterns accurately. Below that it still runs, just with less to learn from.
The attribution model feeds the "Conversions" column, and that column feeds Target CPA, Target ROAS, and ECPC. Switching models is not a reporting change, it is a bidding change, and your targets will need revisiting afterward.
Data-driven is the default for most conversion actions. Many accounts are running it without anyone having decided to, which matters given the point above about bidding.
Google notes that depending on data availability, the two models can return the same result. If your paths are short and single-touch, DDA has nothing to redistribute.
Google Ads has a Model comparison report under Goals then Attribution that shows two models side by side. Compare last click against data-driven on the Cost / conv. and Conv. value / cost columns, and you will see which keywords, ad groups, and campaigns last click has been undervaluing. That is the honest way to size the change before making it. There are also "current model" columns you can add to your reporting, Conversions (current model) and its siblings, which let you see how historical data would have looked under the model you just selected. Use both. Switching attribution models without first measuring the delta is how accounts end up with targets that no longer match the numbers the bidding is now seeing.
Here is the part that matters for anyone spending real money on lead generation. Data-driven attribution is excellent at one job: deciding how to split credit among interactions Google already knows about. It is a redistribution engine. What it cannot do is see anything outside Google's own inventory. It does not know the customer also clicked a Meta ad. It does not know they called instead of filling in a form. And critically, it does not know which of those conversions became actual revenue.
So if the conversion action you feed it is a form fill, DDA will do a sophisticated job of working out which keywords produce the cheapest form fills. It will be more accurate than last click and still aimed at the wrong outcome. The upper-funnel term it just promoted might be the one generating price shoppers. Nothing in the model can tell you, because the model was never shown what happened after the lead.
The fix runs underneath the model, not inside it. Capture the click identifier on arrival, persist it server-side for as long as your sales cycle actually takes, and import the closed outcome back against that click. Now data-driven attribution is redistributing credit across paths that led to revenue rather than paths that led to interest, and the model's intelligence is finally pointed at something worth optimising. Deterministic capture and DDA are not alternatives; DDA gets dramatically better when the conversion underneath it is real.
Why platforms still double-count across channels → · Feeding closed revenue back to Google →
Data-driven attribution is a Google Ads model that assigns conversion credit based on your own account data rather than a fixed rule. It compares the paths of customers who converted against those who did not, identifies the ad interactions that consistently precede conversions, and gives more credit to those. Each model is specific to the individual advertiser, and it covers clicks and video engagements on Search including Shopping, YouTube, Display, and Demand Gen.
All conversion actions are eligible for data-driven attribution regardless of volume, so it will run on any account. Google recommends at least 200 conversions and 2,000 ad interactions in supported networks within a 30-day period for the model to identify patterns accurately. With less data it still functions, but it has fewer paths to learn from and may closely resemble last click.
For deciding how to split credit among Google touchpoints, yes. Last click ignores every interaction before the final one, which systematically undervalues upper-funnel keywords. Data-driven attribution corrects that using your actual conversion paths. It does not, however, see non-Google channels, phone calls outside your tracking, or whether a conversion turned into revenue, so it improves the accounting rather than the underlying measurement.
Only two. First click, linear, time decay, and position-based were deprecated, and conversion actions using them were automatically upgraded to data-driven. Data-driven and last click are the remaining options, and data-driven is the default for most conversion actions.
Yes, and this is the most commonly missed consequence. The attribution model determines how conversions are counted in the "Conversions" column, and automated strategies including Target CPA, Target ROAS, and ECPC optimise against that column. Changing the model changes what the bidding sees, so targets usually need revisiting afterward. Use the Model comparison report to measure the delta before you switch.
Data-driven attribution decides which clicks earn credit. Conversion value rules then multiply that credit by location, audience, or device. Used carelessly, the multipliers distort the signal Smart Bidding learns from. We explain when the rules help and when to leave them off.
Next Step
If you are spending $30,000 or more per month, data-driven attribution is only as useful as the conversion sitting underneath it. We build the deterministic layer that makes the model worth having. Let's talk.
Request a Free PPC Audit →