GA4 Attribution Settings: What to Change and What to Leave
GA4 gives you two attribution decisions: a reporting attribution model and a pair of lookback windows. Both live in one Admin screen (Admin, then Data display, then Events, then Attribution settings, where Google’s own help page sends you). For most accounts the right move is to confirm Data-driven attribution is selected, set the lookback windows to match your real sales cycle, and then stop fiddling. The settings are not where the trouble starts. The trouble starts when people assume they control more than they do.
This guide covers what each setting does, which model deserves your trust, how to pick a window, and why GA4 will never agree with Google Ads. Fair warning: we have opinions about the common habits here, and a couple of them are a waste of your time.
TL;DR
- GA4 now offers three reporting models: Data-driven, Paid and organic last click, and Google paid channels last click. First click, linear, time decay and position-based are gone.
- The model you choose only changes event-scoped reports. Your standard Traffic Acquisition and User Acquisition reports stay on last non-direct click no matter what you pick.
- Lookback windows default to 30 days for acquisition events and 90 days for everything else. Match them to how long your buyers actually take to decide.
- Direct traffic gets credit only when direct is the entire path. Nobody gets to disable that.
- GA4 and Google Ads will not match. Pick one as the source of truth for each question and move on.
Table of Contents
- Where GA4 attribution settings live
- The three attribution models, and which one to pick
- What the model actually changes (and what it doesn’t)
- Lookback windows: set them to your sales cycle
- Direct traffic doesn’t get credit (mostly)
- Why GA4 and Google Ads will never match
- A sane setup in five steps
- Our take: attribution is a reporting question, not a strategy
- When you want someone to own this
- FAQ
- Sources
Where GA4 attribution settings live
Open your property, click Admin, and under Data display choose Events, then Attribution settings. (Older screenshots show a standalone Attribution settings item, which is why half the tutorials you find look slightly wrong.) You’ll need Editor access or higher to change anything, and the settings apply to the property, not to a single report or user.
There are two things on that screen:
- Reporting attribution model. How credit for a key event gets split across touchpoints.
- Key event lookback windows. How far back GA4 looks for touchpoints that can earn that credit.
A word on vocabulary, since it still trips people up. What GA4 used to call conversions are now key events. The reports and the API say “key events” and the settings page does too, so when a stakeholder asks about “conversion attribution settings” they mean this screen.
If you manage many properties, the Admin API exposes the same attribution settings as a singleton resource per property, including the lookback windows and the reporting model. Useful for auditing a client portfolio without clicking through forty accounts.
The three attribution models, and which one to pick
Google’s attribution documentation lists three models available in GA4’s attribution reporting:
| Model | How it assigns credit | Use it when |
|---|---|---|
| Data-driven attribution | Machine learning distributes credit across touchpoints based on your account’s observed paths | Default for nearly everyone, especially if you run Google Ads |
| Paid and organic last click | All credit to the last non-direct click, paid or organic | You want a simple rule you can explain in one sentence |
| Google paid channels last click | All credit to the last Google Ads click when one exists in the path, otherwise last click | You only care about judging Google Ads |
The old rule-based models, first click, linear, time decay and position-based, were removed from all properties in late 2023. If a tutorial still tells you to pick position-based, close the tab. It’s describing a menu that no longer exists.
Data-driven attribution: the default is usually right
Data-driven attribution looks at converting and non-converting paths and assigns fractional credit based on what it observed. Its weakness is that you can’t inspect the arithmetic. As Search Engine Land’s attribution guide notes, Google moved from rigid rules to machine learning, and the price is reduced transparency. A position-based model was easy to defend in a meeting. “The algorithm says so” is harder.
Still, the old rules were arbitrary too. Giving 40 percent to the first touch and 40 percent to the last was a guess dressed up as a policy. We would rather work with an imperfect model learned from your data than a perfect-looking one that was invented.
There’s a catch for smaller accounts. According to TaggingDocs’ write-up of the settings, data-driven models need meaningful volume, on the order of 400 key events and 4,000 path observations in 30 days, to behave reliably. If your site converts a dozen leads a month, you aren’t getting a sophisticated model. You’re getting something that leans on last-click behavior with a fancy label. In that case, don’t agonize. Stay on the default and use your own judgment about the paths, because a model can’t learn from data you don’t have.
Paid and organic last click: the honest, boring option
This one credits the last non-direct touch, from any channel. It’s easy to explain and easy to reconcile with the standard acquisition reports, which already use last non-direct click. If you run a small local business with a short path (someone searches, clicks, calls), the extra machinery of data-driven attribution buys you very little. Last click is fine. Stop apologizing for it.
Google paid channels last click: a narrow tool
This variant hands credit to Google Ads touches when they appear in the path. Its legitimate use is a team that judges Google Ads performance exclusively and wants GA4 to lean in the same direction. Its illegitimate use is making Google Ads look better in a review. Anyone running paid media should be suspicious of a setting that flatters the party that sells the clicks. Our advice: don’t choose it unless you can say in one sentence why Google Ads should win ties.
What the model actually changes (and what it doesn’t)
This is the part nobody warns you about. Changing the reporting attribution model does not rewrite every report. Per Google’s documentation, it applies to reports built on event-scoped traffic dimensions for key events: source, medium, campaign and default channel group in those contexts. The user-scoped and session-scoped dimensions that power your standard Traffic Acquisition and User Acquisition reports remain on last non-direct click.
The consequence is that two reports in the same property can disagree about where a lead came from, and both are correct for their scope. Here’s the practical map:
- Event-scoped (follows your chosen model): key event attribution in the Advertising section and in explorations using event-scoped dimensions.
- Session-scoped (fixed to last non-direct click): Traffic Acquisition.
- User-scoped (fixed to last non-direct click): User Acquisition.
Before you compare any two numbers, ask which scope each report uses. Half the “GA4 is broken” threads we see are really “I compared two different questions.” The same logic explains why conversion tallies diverge across dashboards in general, a theme covered well in this event analytics metrics guide.
Changes apply to historical data too. Because the model is applied at report time rather than baked into collection, flipping it recalculates the past, so you can test a different model without losing history. That makes experimenting low risk, though it also means the number you screenshotted last quarter may not match today’s.
Lookback windows: set them to your sales cycle
The lookback window decides which touchpoints are old enough to be ignored. Google’s settings page gives separate controls for acquisition events and for all other key events. The defaults and options:
| Event type | Default window | Other options |
|---|---|---|
| Acquisition events (first_open, first_visit) | 30 days | 7 days |
| All other key events | 90 days | 30 or 60 days |
The logic is simple and almost nobody applies it: your window should reflect how long your customers actually take to buy.
- Impulse e-commerce, under $100: A 30-day window for other key events is plenty. Longer windows hand credit to touches that had nothing to do with the order.
- Home services and local lead gen: People call within days. Thirty or sixty days is fine.
- B2B, high-ticket, long consideration: Keep 90. A buyer who read your blog in July and booked a demo in September deserves that July touch to count.
The common mistake is shortening the window to “get cleaner data.” You don’t get cleaner data. You get less of it, and the touches you throw away are usually the early, awareness-stage ones that your paid channels don’t take credit for. If you have never measured your own time-to-convert, do that before touching the setting. GA4’s path and conversion reports give you a decent read.
Direct traffic doesn’t get credit (mostly)
Every GA4 model ignores direct visits unless the whole path is direct. That’s by design: someone who clicked a paid ad on Monday and typed your URL on Thursday is credited to the ad. It’s the right call, and it’s the reason your direct numbers look small compared with older analytics tools.
It also means direct traffic is not a measure of brand strength in GA4. If your direct conversion count is climbing, look at your untagged email links, your PDF links and your dark social (messaging apps strip referrers). Those often sit in “direct” or “unassigned” by accident rather than by intent. Fixing your UTM discipline does more for attribution accuracy than any dropdown on the settings page. Our GA4 conversion tracking playbook walks through the tagging side.
Why GA4 and Google Ads will never match
Align the models, align the windows, link the accounts, and the numbers will still differ. Google’s own guidance is to treat them as separate perspectives, because Analytics runs its own processing, its own time zone handling and its own attribution rules. Google Ads counts conversions on the click date or the impression with its own windows, and GA4 counts by the event date inside its model.
So stop trying to reconcile them to the decimal. Decide what each tool is for:
- Google Ads for bidding and in-platform optimization.
- GA4 for cross-channel comparison and what happened on the site.
- Your CRM for what actually became revenue.
That last one is where the real answer sits. Neither tool knows which lead signed a contract, which is why we push clients to feed closed-deal data back into the ad platforms. If you want the full argument, CRM to ads integration is worth ten minutes. And if Google Ads numbers look healthy in the platform but nothing is landing, this three-step fix comes before any attribution tweaking.
A sane setup in five steps
- Confirm your key events are right. Attribution only distributes credit for events you’ve marked as key. A model applied to a junk event list is just a faster way to be wrong.
- Check the reporting model. If it isn’t Data-driven and you have a reason, write that reason down. If you have no reason, switch it.
- Set the lookback windows from your own sales cycle. Not from a blog post, including this one.
- Link Google Ads and Search Console, and make sure auto-tagging is on, so paid clicks are classified correctly instead of drifting into “Unassigned.”
- Document the choice in a note on the property or in your reporting doc, with the date. Settings changes recalculate history, so future-you needs to know why the trend line bent.
Then leave it alone for a quarter. Changing models every month to find the flattering number is not analysis. It’s shopping.
Our take: attribution is a reporting question, not a strategy
We run paid and organic programs for businesses that have outgrown their agency, and the pattern is consistent: the teams that obsess over attribution settings are usually the ones whose underlying measurement is shaky. They are debating whether credit goes 60/40 or 50/50 while half their form fills are untracked and their CRM never talks to their ad accounts.
Here’s what we believe, and how we work as a result:
- The model matters less than the plumbing. Clean key events, disciplined UTMs and a connected CRM beat any model choice.
- Reports you can’t check are reports you can’t trust. That’s why our reporting runs on live dashboards tied to GA4 and Google Business Profile, not a monthly PDF that shows up after the budget is already spent.
- Spend belongs on creative and execution, not process. An agency that burns a month debating attribution philosophy is spending your money on the wrong thing.
- Pick a model, state it, and be consistent. Anything else invites quiet cherry-picking.
If your agency can’t tell you which attribution scope its monthly numbers come from, that’s your answer about how closely anyone is watching.
— Chris Breikss
When you want someone to own this
If you’d rather have the measurement set up correctly and the campaigns running against it, that’s what we do. Our live marketing dashboard shows how GA4 data looks when you can check it anytime, and our Google Ads and LSA management is built on the same reporting. See more on dashboards and reporting, or start from Rivetline.
FAQ
Does changing the GA4 attribution model change all my reports?
No. It affects reports that use event-scoped traffic dimensions for key events. Standard Traffic Acquisition and User Acquisition reports remain on last non-direct click regardless of your selection. If two reports disagree, check their scope first.
What lookback window should I use in GA4?
The defaults are 30 days for acquisition events and 90 days for other key events. Keep 90 if your buyers take weeks to decide, as in B2B or high-ticket services. Drop to 30 only if your customers reliably convert within a few weeks.
Is data-driven attribution better than last click?
Usually, if you have enough conversion volume for it to learn from. It spreads credit across the path based on observed behavior, but you can’t inspect its math. Low-volume sites get little extra value and can safely treat last click as the working rule.
What happened to first click, linear and position-based models in GA4?
Google removed them from all properties in late 2023. Today you choose among Data-driven, Paid and organic last click, and Google paid channels last click. Older tutorials that mention the removed models are out of date.
Why don’t my GA4 and Google Ads conversions match?
The two tools use their own processing, time zones and attribution rules, so some gap is normal even with aligned settings. Use Google Ads for bidding decisions, GA4 for cross-channel views, and your CRM for revenue.
Sources
- Select attribution settings, Analytics Help
- Get started with attribution, Analytics Help
- AttributionSettings, Google Analytics Admin API
- Your guide to Google Analytics 4 attribution, Search Engine Land
- Attribution Settings, TaggingDocs
- October 2023 GA4 & Google Updates, Napkyn
