When Google moved everyone off last-click, whose budgets actually changed?

When Google moved everyone off last-click, whose budgets actually changed?

In 2021-2023 Google Ads deprecated rules-based attribution and defaulted accounts to data-driven attribution. Millions of advertisers had their credit reshuffled without changing a single ad. The question: what systematically moved, and why?

The change. Data-driven attribution uses a model (Google's is a Shapley-value-style approach trained on each account's converting and non-converting paths) to distribute credit by each touchpoint's estimated contribution, rather than a fixed rule like last-click. It learns from the account's own journey data.

What advertisers observed. Aggregated reports across agencies documented a consistent pattern after the switch: upper- and mid-funnel keywords and generic search gained credit, while branded and final-click terms lost 10-30%. Video and Discovery campaigns, previously near-invisible under last-click, showed measurable contribution for the first time. Total conversions were unchanged; their distribution shifted toward assist touches.

The nuance. Data-driven attribution corrects last-click's position bias — it stops the final touch from hoarding credit. But it remains a credit-allocation model over observed paths, not a causal measurement. It will still credit a touch inside journeys that would have converted regardless. Smart Bidding then optimizes to those modeled credits, which can entrench the model's assumptions.

Bottom line for practitioners: the data-driven switch likely made your reports more accurate about position and less honest about causation — better than last-click, still not an experiment. When the model reallocates budget toward assist channels, validate the biggest movers with a holdout before letting automated bidding compound the shift.