Attribution models determine which clicks along a user's conversion path receive credit for the conversion. A user who sees your ad three times — searches broadly, clicks a competitor comparison ad, then converts on a branded search — generates three touchpoints. Which one gets the credit in your reporting depends entirely on which attribution model you are using. And which model you use determines which campaigns Google's Smart Bidding optimizes toward.
Google simplified its attribution options in 2023 by removing linear, time decay, and position-based models. The remaining models are: last click, first click, and data-driven attribution. For most accounts, this simplification is actually the right move — the removed models had theoretical appeal but produced noisy signals in practice. Understanding what the remaining three models do and which accounts they suit is now straightforward.
Last click attribution assigns 100% of the conversion credit to the final click before the conversion. A user who clicked a Display ad, then a Search ad for a generic keyword, then a branded Search ad before purchasing — the branded Search ad gets all the credit. Last click attribution consistently overstates the value of bottom-of-funnel keywords and branded terms, and understates the value of awareness and consideration campaigns. If your account runs both prospecting and conversion campaigns, last click will make your prospecting campaigns look inefficient relative to their actual contribution.
The practical case for last click: it is simple, transparent, and directly measurable. When you optimize toward last-click conversions, you are optimizing toward actions that directly preceded purchases. For accounts with simple, single-step sales cycles — a user searches, clicks, buys in a single session — last click accurately reflects reality because there is no multi-touch path to attribute. For product categories with long consideration cycles or multiple research steps, last click distorts the picture.
First click attribution assigns 100% of the credit to the first touchpoint — the ad that introduced the user to your product. This model systematically overvalues awareness campaigns and undervalues the campaigns that close conversions. It is the mirror image of last click. In practice, first click is rarely the right choice for performance optimization because you are optimizing toward the campaigns that generate exposure, not the ones that generate revenue.
Data-driven attribution (DDA) uses machine learning to distribute conversion credit across all touchpoints based on their actual contribution to the conversion path. Google's model analyzes the paths that led to conversions versus paths that did not, and assigns fractional credit to each touchpoint based on its incremental impact. A click that appears frequently on converting paths but rarely on non-converting paths receives high credit. A click that appears equally on both paths receives low credit.
DDA is the most accurate model when it works — and 'when it works' is the key qualifier. Google requires a minimum of 300 conversions per month with sufficient path diversity to train the model. Below this threshold, the model does not have enough data to produce reliable credit assignments and falls back to a less accurate estimation. Accounts with 50–150 conversions per month that select DDA are often running a model that is not meaningfully better than last click but adds opacity to their reporting.
If you meet the 300-conversion threshold, DDA is almost always the right choice for performance campaigns. It typically shifts budget toward mid-funnel keywords and earlier touchpoints that last click undervalues, improving the efficiency of the full acquisition funnel. Accounts that switch from last click to DDA with sufficient conversion volume frequently see reported CPA increase slightly (because conversions are now attributed more accurately across more campaigns) while actual conversion volume and revenue hold steady or improve — the reporting looked better under last click because branded terms were taking undeserved credit.
For accounts below the DDA threshold, last click is the pragmatic choice. It is transparent and easy to explain to stakeholders. The distortion it creates — overvaluing bottom-funnel terms — is predictable and can be managed by keeping prospecting and conversion campaigns structurally separate and evaluating prospecting campaigns on view-through conversions and assisted conversion data rather than last-click conversions.
Attribution model affects Smart Bidding directly. When you use Target CPA or Target ROAS, Google's algorithm optimizes toward the conversion signals defined by your chosen model. If you use last click attribution and set a €30 Target CPA, the algorithm treats the last-click conversion as the signal to optimize toward. Switch to DDA and the algorithm receives fractional signals from multiple touchpoints, which can meaningfully change how it allocates budget across campaigns. Always audit your attribution model before making significant Smart Bidding changes — the model and the bidding strategy need to be aligned. Digital Face surfaces your current attribution setup, flags accounts where attribution model and bidding strategy may be misaligned, and identifies campaigns where model selection is likely distorting reported CPA. Free plan at digital-face.nl, no credit card required.