Last click vs data-driven attribution, which is better?

Usually data-driven, if your store has enough journeys with more than one click. It spreads credit across the touches GA4 or Google Ads recorded, while last click gives it all to the final one. If most sales follow a single visit, the two agree. Neither one tells you what a channel caused.

What one store's data shows

The usual answer says data-driven is smarter, so switch and stop worrying. One store's export puts a different question first. Where could the two models disagree at all?

One store's anonymised GA4 export, 1 January 2024 to 21 August 2026. It holds shares of revenue only: no ad spend, no order counts.

Start with the Journeys sheet. Journeys with 1 touch hold 79.5% of revenue on the Journeys sheet, with 0.5 days to buy. On a one-touch path, last click and data-driven return the identical answer. So in this store, the two models can only disagree about revenue outside that 79.5%.

The Channels sheet adds a useful limit. It shows each channel three ways, and the touched view gives a channel full credit for every journey it appeared in. That makes touched a ceiling. No model that splits credit inside a path can hand a channel more than its touched share.

Direct holds 57.7% of revenue on the Channels sheet under last click, and the same 57.7% at its touched ceiling. First click matches too. Last click already sits at that ceiling, so data-driven has no room to give Direct more in this store.

Paid Social sits at 0.0% on the Channels sheet in all three views, ceiling included. A model can only share credit among channels on a recorded path, and by revenue, Paid Social barely appears on any here. The export cannot say whether this store ran social ads at all. If you do, and your own export shows that row, the fix is tagging, not a cleverer model.

So for this one store, the model choice is an argument about a minority of the money. The export cannot say which model is closer to the truth on that minority. It can show how little rides on the answer here.

Why does the usual answer mislead?

Because data-driven sounds like measurement. It is still a way to split credit for sales that already happened, among touches GA4 or Google Ads recorded.

GA4's help calls the model counterfactual. Its technical note goes further: the models compute the counterfactual gains of Google ad exposures by training on data from randomized controlled trials. That is closer to cause than any rule. But that training concerns Google's ads, and the page says nothing similar about your emails or your Meta clicks.

Switching models moves credit, not sales. Google's own help has an example. A switch to data-driven takes a Brand campaign from 200 conversions to 150, and a Generic campaign from 50 to 100. The total stays put; only the hero changes.

In Google Ads, that shift reaches your bids. Google says the model you select affects how automated bid strategies such as Target ROAS optimise. Switch without resetting targets, and Google warns the change in credit can cause over- or under-bidding.

Last click has a virtue people skip: you can see exactly how it errs. It overpays whatever closes the sale, such as brand search, email and retargeting. Data-driven errs too, but you cannot inspect how it sets its weights. One bias you can correct for. The other you take on trust.

What can neither model tell you?

Whether a sale needed the ad. Both models share out credit for sales that happened. To learn what a channel adds, switch it off in some regions and compare total sales there.

What never reached a path. A channel with no recorded touches gets nothing from either model, which is about where the Paid Social row above sits. A view on Instagram, a podcast mention or a shop shelf leaves no touch to weigh.

Whether a channel pays. Credit is not margin. You still need spend and margin to know whether a channel earns its keep, under either model.

What to do this week

1. Compare purchases only. In GA4, open Advertising > Attribution > Attribution models and pick purchase in the key events drop-down. By default, every key event is selected and added together. Pass: the table shows purchase revenue alone. Fail: sign-ups or other key events are mixed in, so the % Change columns mean little.
2. Look for paid channels with no row. In the same report, set one Attribution model (non-direct) column to Data-driven and the other to Paid and organic last click. Then list every channel you pay for. Pass: each has purchase revenue under both models. Fail: one is missing under both, so its clicks are likely arriving untagged or mislabelled, and no model can fix that.
3. Filter to the campaign you plan to cut. Click Add filter, create an Include filter, select User campaign under Acquisition and pick the campaign. Pass: it earns similar revenue under both models. Fail: it gains a lot under data-driven, so last click is underpaying it; test it before you cut.

Check the homework. Your GA4 Attribution paths export already holds the evidence. Causality Engine reads that one file and shows what each channel caused next to what last-click gave it, in 1 to 2 minutes, for €99 once (excluding VAT), refundable within 30 days. Check the homework

Sources, 1 October 2026: Get started with attribution (Google); Key event attribution models report (Google); Select attribution settings (Google); About attribution models (Google); Best practices for managing attribution model changes (Google)

Related answers

• All Attribution models questions
• How to test data-driven against last click before switching
• Last click or data-driven for a food and beverage store?
• Last click vs. data-driven attribution
• Causal attribution, explained