Attribution Settings
Attribution settings in GA4 determine which touchpoint gets credit for a key event and how far back GA4 looks for touchpoints. These settings matter most if you use the Advertising section of GA4 or export attribution data to Google Ads for bidding.
Accessing attribution settings
Section titled “Accessing attribution settings”Go to Admin → Attribution Settings (under the Property column).
You will see:
- Reporting attribution model — the default model for the Advertising section
- Lookback windows — how far back to attribute key events
Attribution models
Section titled “Attribution models”As of November 2023, GA4 offers two attribution models in the UI:
Data-driven attribution
Section titled “Data-driven attribution”Uses machine learning to assign fractional credit to each touchpoint based on its actual contribution to key event. The model analyzes patterns across converting and non-converting paths to estimate the incremental value of each channel.
Requirements: Your property needs sufficient data for the model to train. Google recommends at least 400 key events and 4,000 conversion path observations over a 30-day period. Below this threshold, the model falls back to last-click.
Best for: Properties with high key event volume and diversified channel mix. Data-driven attribution is the default for all new GA4 properties and the best choice if you have sufficient data.
Last click
Section titled “Last click”100% of credit goes to the last touchpoint before the key event. Direct traffic is excluded — if the last non-direct touchpoint was Paid Search and the user came back directly to convert, Paid Search gets the credit.
Best for: Simple implementations, low key event volume, or when you want a baseline that is easy to explain to stakeholders.
Deprecated models
Section titled “Deprecated models”The following attribution models were removed from GA4 in November 2023: First click, Linear, Time decay, and Position-based (U-shaped). These models are no longer available in the UI.
To perform multi-touch attribution analysis, use one of these approaches:
- BigQuery SQL: Build custom multi-touch attribution queries using raw event data
- GA4 BI Connector: Use Looker Studio with the BI Connector to implement custom attribution models
- Third-party tools: Connect GA4 to your marketing mix modeling platform or attribution software
If you have specific multi-touch attribution needs that previously relied on these models, consider implementing the attribution logic in BigQuery or exporting data to a dedicated attribution platform.
Attribution changes are retroactive
Section titled “Attribution changes are retroactive”This is the most important thing to know about GA4 attribution settings: changing your attribution model retroactively restates historical data in the Advertising reports.
If you switch from last-click to data-driven today, your historical channel performance in the Advertising section changes to reflect data-driven attribution. This is intentional — you can compare models on the same historical data — but it surprises analytics practitioners used to Universal Analytics, where model changes only affected future data.
The attribution model setting does not affect:
- Standard acquisition reports (Traffic acquisition, User acquisition) — these always use last non-direct click
- Raw event data in BigQuery
- Key events counted in the Events report
It only affects the Advertising section’s key event attribution.
Lookback windows
Section titled “Lookback windows”Lookback windows define how far back GA4 looks for touchpoints when attributing a key event.
| Key event type | Default window | Range |
|---|---|---|
| Acquisition key events (first_visit, first_open) | 30 days | 7, 14, 30, 60, 90 days |
| All other key events | 90 days | 7, 14, 30, 60, 90 days |
Shorter window — fewer touchpoints considered, credit concentrates on recent channels, shorter-path key events represented more accurately.
Longer window — more touchpoints considered, credit spreads across older channels, long sales cycles represented more accurately.
Set the lookback window to match your actual key event cycle. If your typical path from first visit to key event is 3-7 days, a 30-day window is generous. If your B2B sales cycle averages 45 days, you may need a 90-day window.
Which reports use attribution settings
Section titled “Which reports use attribution settings”The attribution model and lookback windows affect:
Advertising section:
- Attribution: Model comparison
- Attribution: Conversion paths
- Attribution: Top conversion paths
Google Ads imported key events: When you import GA4 key events to Google Ads, those key events use the GA4 attribution model for optimization.
Standard reports (NOT affected):
- Reports → Acquisition → Traffic acquisition
- Reports → Acquisition → User acquisition
These reports always use last non-direct click, regardless of your attribution settings.
Comparing models
Section titled “Comparing models”The Attribution → Model comparison report lets you see how different models distribute credit across channels without committing to a model change. Since only Data-driven and Last-click remain, this comparison is useful if you want to see the difference before enabling Data-driven:
- Go to Advertising → Attribution → Model comparison
- Select Data-driven and Last-click to compare
- The table shows key events and key event value attributed to each channel under each model
This is useful for understanding the impact of switching to data-driven attribution before making the change.
Data-driven attribution and consent mode
Section titled “Data-driven attribution and consent mode”Data-driven attribution requires sufficient data to train its model. Consent mode can affect this:
- When users decline analytics cookies, GA4 collects aggregate, non-identifying behavioral data
- This modeled data contributes to the overall dataset but at lower confidence
- For properties with high opt-out rates (common in EU), data-driven attribution may have lower accuracy
In consent-restricted environments, last-click attribution is often more reliable because it depends only on the single observed touchpoint rather than pattern matching across the full user journey.
Common mistakes
Section titled “Common mistakes”Changing attribution models without notifying stakeholders
Section titled “Changing attribution models without notifying stakeholders”Attribution model changes restate historical data. A monthly channel performance report will show different numbers for past periods after the change. Always communicate attribution changes in advance and document them with a note in your reporting.
Using the wrong lookback window for your sales cycle
Section titled “Using the wrong lookback window for your sales cycle”A 7-day lookback window on a B2B SaaS product with a 60-day trial period means most early-funnel touchpoints get zero attribution credit. The key event appears to come entirely from “last click” within the window, making email nurture campaigns and content marketing look worthless.
Assuming standard reports use your attribution model
Section titled “Assuming standard reports use your attribution model”Many analysts configure data-driven attribution and then wonder why the Traffic acquisition report looks unchanged. Standard reports use last non-direct click. Your attribution model settings only apply to the Advertising section. This is correct behavior, not a bug.
Not understanding data-driven requirements
Section titled “Not understanding data-driven requirements”Data-driven attribution silently falls back to last-click when key event volume is too low. If your property has 50 key events per month, data-driven is not operating — you are getting last-click with a “data-driven” label. Check the model comparison to see if data-driven is actually distributing credit differently than last-click; if the numbers are identical, the model has not trained.