How does it decide who gets the credit?
By comparing paths. Google looks at the sequences of ad interactions that ended in a conversion and the ones that did not, and gives more credit to the interactions that turn up more often in the converting paths (Google Ads Help, About data-driven attribution). An ad that people tend to click early, before the paths that end in a sale, earns a share even if it was never the last click.
Google announced in 2021 that data-driven attribution would become the default for new conversion actions, and dropped the minimum data it used to require. In 2023 it retired the other rules-based models except last click, so for most accounts the choice today is between the two.
What can it learn from a small account?
Less than the name suggests. Google's help page says the model works with any volume of conversions, and recommends at least 200 conversions and 2,000 ad interactions in 30 days for it to assign credit more precisely. Many owner-led businesses get 20 or 30 conversions a month.
There is also a simpler limit. Data-driven attribution can only redistribute credit when a customer clicked more than one ad. In a small account most buyers click one ad and convert, and for those buyers every model gives the same answer. So in practice the model changes the credit on a handful of conversions, and its view of those few is based on very little data. It is not harmful, but its reports should be read as a rough split.
How different are the two models for a Canberra driving school? (example figures)
A driving school in Canberra booked 25 lesson packages from Google Ads last month. Nineteen of those learners (or their parents) clicked one ad and booked. Six clicked two or more ads first. The owner compares the credit under each model.
| Campaign | Last click | Data-driven |
|---|---|---|
| Searches for the school's name | 9 | 7.6 |
| "Driving lessons Canberra" | 11 | 12.1 |
| Suburb campaigns (Belconnen, Tuggeranong) | 5 | 5.3 |
| Total | 25 | 25 |
The two models disagree on less than 2 of the 25 bookings. Data-driven moves a little credit away from searches for the school's name, which people often click last after finding the school some other way, and towards the generic search that introduced them. That direction is plausible and useful. The size of the shift, on six multi-click paths, is too small to act on by itself, and a different six next month could move it the other way.
For this school, the sensible course is to leave the default in place and judge campaigns over a quarter.
Where does data-driven attribution touch CoraVision?
CoraVision reads conversions and conversion value from Google Ads every night, and those figures arrive already credited by whichever model each conversion action uses, which for most accounts is data-driven. CoraVision also reads the bidding type of each campaign, and Smart Bidding learns from the same credited conversions, so a model change shows up in both. For a small account, that split is best read over a quarter, as the driving school does.
Questions
Can I switch a conversion from data-driven to last click?
Yes. In Google Ads the model is set on each conversion action, in its settings. Changing it alters how credit is split from then on and what Smart Bidding learns from, so choose once and leave it, or the before and after figures stop being comparable.
Why does my report show 7.6 conversions?
Because data-driven attribution can split one conversion between several ads. If a customer clicked a brand ad and a generic ad, each might get part of that one conversion. Add the fractions across all your campaigns and they come back to whole numbers.
Does Google Ads' data-driven model give credit to Meta or email?
No. Google Ads shares credit only among Google ads the person interacted with. Google Analytics 4 has its own data-driven model that looks across channels it can track, such as paid social and tagged email, so the two can tell different stories about the same sale.