GA4 quietly changed how attribution works compared to Universal Analytics. The default model is now “data-driven” instead of last click, the attribution settings are buried in the Admin panel, and most marketers never touch them. That means your conversion reports might be splitting credit across channels in ways you don’t expect — or want.
This guide covers every attribution setting in GA4, what each one does to your data, and which configuration makes sense for different business types.
Where to Find Attribution Settings
GA4 Admin → Property → Attribution settings
There are three things you can configure:
- Reporting attribution model — how credit is distributed in reports
- Lookback window — how far back GA4 looks for touchpoints
- Channels eligible for attribution — whether Google paid channels get priority
Let’s break down each one.
The Attribution Models
Data-Driven Attribution (Default)
GA4’s data-driven attribution (DDA) uses machine learning to distribute credit based on your actual conversion data. It looks at the paths that lead to conversions versus the paths that don’t, and assigns credit proportionally.
Example: If your data shows that users who see an Instagram ad AND then click a Google ad convert at 3x the rate of users who only click Google ads, DDA gives Instagram partial credit.
| Pros | Cons |
|---|---|
| Uses your real data, not a formula | Requires volume (needs enough conversions to model) |
| Adapts as behavior changes | Black box — you can’t see why credit is assigned |
| Accounts for channel interactions | Can be inconsistent at low volumes |
| Google’s recommended model | May shift credit unexpectedly |
When to use it: You have 300+ conversions per month and want the most “accurate” attribution based on actual patterns.
When to avoid it: Low conversion volume (under 100/month), or you need predictable, explainable attribution for stakeholders who won’t accept “the algorithm decided.”
Paid and Organic Last Click
Credit goes to the last click before conversion, excluding direct. If the last touchpoint was a direct visit but the previous touchpoint was Google Ads, Google Ads gets credit.
Example: User clicks a Meta ad → visits directly 3 days later → converts. Meta gets credit.
When to use it: You want simple, defensible attribution. Common for businesses reporting to clients or executives who want straightforward channel ROI.
Google Paid Channels Last Click
Same as last click, but Google Ads always gets credit if it was anywhere in the path within the lookback window.
When to use it: Rarely recommended. This model is biased toward Google’s own channels. It exists for backward compatibility with older Universal Analytics setups.
When to avoid it: Almost always. It systematically over-credits Google paid channels and under-credits everything else.
How Models Change Your Numbers
Here’s the same data viewed through each model:
Scenario: User sees a Meta ad (click), then an organic search, then a Google Shopping ad (click), then converts.
| Model | Meta Ads Credit | Organic Search | Google Shopping |
|---|---|---|---|
| Data-Driven | 30% ($30) | 20% ($20) | 50% ($50) |
| Last Click | 0% | 0% | 100% ($100) |
| Google Paid Last Click | 0% | 0% | 100% ($100) |
With data-driven, Meta and organic get partial credit. With last click, Google Shopping gets everything. Same data, dramatically different budget decisions.
This is why your Meta Ads manager and your Google Ads manager will always argue about attribution — they’re looking at models that tell different stories. For a deeper comparison of models across platforms, see our multi-touch attribution guide.
Lookback Windows
The lookback window determines how far back GA4 looks for touchpoints that should receive credit.
Acquisition Conversion Events
For first-visit conversions (like first_open or first_visit events):
- 30 days (default)
- 7 days
This controls how long GA4 looks back for the ad click or touchpoint that brought the user to your site for the first time.
All Other Conversion Events
For returning-visit conversions (like purchase or generate_lead):
- 30 days
- 60 days (default)
- 90 days
How to choose:
| Business Type | Recommended Window | Why |
|---|---|---|
| Impulse purchases (cheap products) | 30 days | Short consideration cycle |
| Ecommerce (mid-price) | 60 days (default) | Standard buying window |
| B2B / High-ticket | 90 days | Long sales cycles |
| SaaS free trial → paid | 90 days | Trial + onboarding period |
The trade-off: Longer windows give more credit to upper-funnel touchpoints (that first blog visit 80 days ago). Shorter windows concentrate credit on lower-funnel touchpoints (the retargeting ad they clicked yesterday).
If you have long data retention settings in GA4, your historical data supports longer lookback windows. But the attribution window and retention window are separate settings — retention determines how long raw data is stored, while lookback determines how far back attribution looks per conversion.
Engaged-View Lookback Window
For video ads (YouTube):
- Disabled
- 3 days
- 7 days (default)
This controls whether YouTube “engaged views” (watched 10+ seconds of a skippable ad without clicking) receive attribution credit.
Recommendation: Keep at 7 days if you run YouTube ads. Disable if you don’t.
Configuration Walkthrough
Step 1: Access Attribution Settings
- Open GA4
- Admin → Property → Attribution settings
- You’ll see three sections
Step 2: Select Your Model
For most businesses: stick with Data-Driven Attribution.
It’s Google’s default and recommended model. Unless you have a specific reason to change it (low volume, stakeholder requirements), DDA gives the most nuanced view.
If you change the model, it applies retroactively to all historical data. This is different from Universal Analytics where model changes only affected future data. In GA4, switching from last click to data-driven will change your past reports too.
Step 3: Set Lookback Windows
Match the window to your sales cycle:
- Short cycle (< 2 weeks): 30-day acquisition, 30-day other
- Medium cycle (2-8 weeks): 30-day acquisition, 60-day other (default)
- Long cycle (> 8 weeks): 30-day acquisition, 90-day other
Step 4: Configure Channel Eligibility
Under “Channels eligible for attribution,” you have two options:
- Google paid channels, organic channels, direct — Includes Google Ads clicks in attribution. This is the standard setting.
- Google paid channels only — Direct and organic are excluded. Only use this if you exclusively care about paid performance.
Recommendation: Keep the default (all channels). Excluding organic and direct from attribution inflates paid channel numbers artificially.
Model Comparison Report
GA4 has a built-in tool to compare how different models attribute your conversions:
- Advertising → Attribution → Model comparison
- Select two models to compare side-by-side
- Filter by conversion event (e.g., just purchases)
This shows you exactly how credit shifts between channels under each model. If Google Shopping goes from 150 conversions (last click) to 120 conversions (data-driven), the missing 30 were shared with other channels that contributed to those same purchases.
Common Questions
”Should I match GA4’s model to Google Ads?”
No. Google Ads has its own attribution model (also data-driven by default). GA4 and Google Ads will always show different numbers because they track differently. Trying to match them is a losing game — see our GA4 vs Google Ads conversion mismatch guide for why.
Use GA4 for cross-channel reporting. Use Google Ads for in-platform optimization. Accept that the numbers won’t match.
”Does changing the model affect Google Ads bidding?”
No. Google Ads bidding uses its own attribution model, not GA4’s. Changing GA4’s model only affects GA4 reports. The exception: if you import GA4 conversions into Google Ads for bidding, then GA4’s model determines the conversion value Google Ads sees.
”My data-driven model keeps changing credit distribution. Is that normal?”
Yes. DDA recalculates continuously as new conversion data comes in. If your channel mix changes (you pause Meta ads, for example), the model will redistribute credit based on the new patterns. This is a feature, not a bug — but it can make month-over-month comparisons noisy.
”I only have 50 conversions a month. Which model should I use?”
At low volumes, data-driven attribution doesn’t have enough data to model reliably. Use Paid and Organic Last Click instead. It’s simple, predictable, and doesn’t require statistical significance to work.
”We report to clients monthly. What should we use?”
Paid and Organic Last Click for client reporting. It’s the easiest to explain (“the last channel they clicked before buying gets credit”) and produces consistent numbers. Use data-driven internally for your own optimization decisions.
Recommended Configurations
Ecommerce (DTC)
| Setting | Value |
|---|---|
| Model | Data-Driven |
| Acquisition window | 30 days |
| Other events window | 60 days |
| Engaged-view window | 7 days (if running YouTube) |
| Channel eligibility | All channels |
B2B / Lead Gen
| Setting | Value |
|---|---|
| Model | Data-Driven (if 300+ leads/month) or Last Click |
| Acquisition window | 30 days |
| Other events window | 90 days |
| Engaged-view window | Disabled |
| Channel eligibility | All channels |
Small Business / Low Volume
| Setting | Value |
|---|---|
| Model | Paid and Organic Last Click |
| Acquisition window | 30 days |
| Other events window | 30 days |
| Engaged-view window | Disabled |
| Channel eligibility | All channels |
Checklist
- Attribution settings reviewed in GA4 Admin
- Reporting model selected based on conversion volume
- Lookback windows match your sales cycle
- Engaged-view window configured for YouTube (or disabled)
- Channel eligibility includes all channels (not Google-only)
- Model comparison report reviewed for current channel credit
- Team aligned on which attribution model is “source of truth”
Attribution settings determine how your marketing budget looks in reports. The wrong model can make a profitable channel look worthless or a wasteful channel look essential. Take ten minutes to configure them correctly.
Want to verify your GA4 is configured correctly — attribution settings, conversion events, and all? Run a free scan to check.