How to Fix Last-Click Attribution in GA4: Settings, Limits
By Joris van Huët
Enterprise Interim CMO & Marketing Leader · 15+ years · 50+ orgs
Published
2026-10-27
The short answer: switch the attribution model away from last click, and the session reports still read last click. You are not misreading it. GA4 has offered three reporting models since November 2023 (data-driven, paid and organic last click, and Google paid channels last click). Switching between them re-splits credit across the same total, per Google's attribution settings page. Set the model, lookback and eligible channels deliberately. Accept that session reports stay last-click whatever you pick. Then read caused credit beside last-click. For that last part, Causality Engine rebuilds last-click from your own GA4 export and puts each channel's caused share next to it. The first finding is free on the start page, and the full read is €99 once.
Disclosure: I founded Causality Engine and own it, and this page recommends it. Every claim about it links to Causality Engine's own pages.
What the settings page moves, and what it leaves alone:
- It moves credit, not sales. For example, a €120 order arrives after Paid Social, then Email, then Organic Search. Last click gives all €120 to Organic Search. A data-driven split can give €40, €30 and €50 instead. Either way the three amounts add up to €120. No model on the settings page asks whether the order happens without Paid Social.
- It has three dials. Reporting attribution model, channels that can receive credit, and the key event lookback window. All three sit under Admin, Data display, Events, Attribution settings, for users with the Marketer role or above, per the settings page.
- The check comes after. Compare models on the same paths report, then set a causal read next to them. Step 7 covers both.
Before you start
- In: Marketer access to the GA4 property, a purchase key event, and the date your Google Ads account was linked.
- Out: settings you chose on purpose, a clear map of which reports follow them, and a last-click column with caused credit beside it.
Step 1: Read what is set today
Open Attribution settings and note the current model, lookback and channel setting. Google's page says a model change applies to historical and future data. It shows up in every key event report that uses event-scoped dimensions such as Source, Medium, Campaign and Default channel group. User- and session-scoped dimensions, such as Session source and First user medium, are not affected.
Step 2: Choose the reporting model
Data-driven is the default for the event-scoped channel dimension, per the default channel group table. And it is a serious model. Google's attribution overview says it evaluates converting and non-converting paths and uses a counterfactual approach, trained on randomised experiments for Google ad exposures. That is more than a position rule does. It is still a split of credit for key events that happened: fractional credits for one key event sum to 1.0, per the paths report page.
Step 3: Set the lookback to your buying cycle
For key events other than first visits and first opens, the lookback is 90 days by default, with 30 or 60 available. Acquisition key events default to 30 days, with 7 as the alternative, per the settings page. Changes apply going forward only. Check the Days to key event column in the paths report first. If most buyers convert inside 30 days, a 90-day window changes little. If many take longer, a shorter window cuts the channels that started those journeys.
Step 4: Decide which channels can receive credit
For GA4's own reports, Paid and organic is the default. For Google Ads reports, Google paid is the default. Switching applies going forward and to every linked Google Ads account, and it can affect bidding there, with a few days' delay, per the settings page. Tell whoever runs Google Ads before you touch it.
Step 5: Know which reports ignore all of the above
Traffic acquisition reports on session source and medium, per Google's UTM page. The session default channel group always uses paid and organic last click, per the channel group page. So after a switch to data-driven, Traffic acquisition and the Advertising reports can show different revenue for the same channel. Neither is broken. They answer differently scoped questions.
Step 6: Compare models on the paths report
Open Advertising, then Key event attribution paths. The chart at the top uses paid and organic last click by default and has a model drop-down. It splits each path into early, mid and late touchpoints (the first 25%, the middle 50% and the last 25%), per Google's paths page. A channel that is large early and small late is the one last-click shortchanges. Note it. That is the channel to watch in step 7.
Step 7: Put caused credit next to last-click
Download the same report as a CSV and drop it on Causality Engine's start page. The privacy policy says the free preview reads the file in your browser and does not upload it. The full read rebuilds last-click from the same file and shows the engine's caused share beside it. Each channel gets a data-health score from 0 to 100 and a next step, per how it works. The GA4 integration page notes that the read is independent of the model you chose in step 2.
Direct matters here. In GA4, Direct keeps credit only when every touchpoint on a path is direct, per the attribution overview. The read splits that remaining Direct credit back to the channels that sent those buyers. The start page shows one store's read, donated with consent (€905,898 of purchase revenue, 1 January to 13 October 2026). In it, 61.67% of Direct's credit belonged to Paid Shopping, Organic Search and Cross-network, per the start page. That is the company's own published example, not a forecast for your store.
Mistakes Google's pages warn about
- Reading old dates under a new model. Data-driven covers paid and organic channels from 1 November 2021 and the rules-based models from 14 June 2021, so earlier ranges show partial data, per the settings page.
- Expecting a lookback change to rewrite history. It applies going forward.
- Hunting for first click or linear. They were removed in November 2023, alongside time decay and position-based.
For Google Ads reports, the out-of-the-box setting lets only Google's own paid channels receive conversion credit. Until someone changes it, the account that bills for the click is the only candidate that report considers for the sale.
Settings first, then the read
Steps 1 to 6 are free and worth doing even if you never buy anything. They make GA4's credit deliberate instead of inherited. They will not tell you what a channel caused. Causality Engine, which I founded, does that from the same export. The first finding is free on the start page, then the read is €99 once, refundable within 30 days, on the pricing page (checked on 27 October 2026). The method is laid out on the last-click attribution page, and the Causality Engine entry covers the product.
FAQ
Can I bring back first-click or linear attribution in GA4?
No. Google removed first click, linear, time decay and position-based in November 2023, per the settings page. For a first-touch view, use the First user default channel group dimension, which is user-scoped and unaffected by the model setting.
Why does Traffic acquisition still look like last-click after I switched to data-driven?
Because it reports session-scoped dimensions, which the model setting does not touch. The session default channel group runs on paid and organic last click, per Google's channel group page.
Does the causal read change if I change the GA4 model?
No. Causality Engine's GA4 integration page describes the read as independent of GA4's attribution model selection: it works from the paths in the export and rebuilds last-click itself.
ABOUT THE AUTHOR
I am Joris van Huët, an enterprise interim CMO. 15+ years, 50+ organizations, including ING, P&G, Nestlé, BNP Paribas, WeTransfer and Vinted. I take on innovation projects (venture building, design sprints), agentic marketing (setting up and running AI agents) and hands-on multi-channel work. See the track record.
I wrote and published this with AI assistance, and I answer for it. Claims about my own experience are limited to the track record above, and a statistic links to its source or is labelled as an example. I sell interim and fractional CMO work, which is why this site exists. How this site is written.
A senior operator doing the work: campaigns, channels, CRM and AI agents, live from day one.