Which attribution model does GA4 use, and should I change it?
| Stores | 16 |
|---|---|
| GA4 recorded (median) | 76% of real orders |
| Middle half of stores | 56% to 89% |
| Stores where GA4 saw under 80% | 9 of 16 |
| Stores where GA4 saw over 110% | 0 |
Stores with at least 20 orders in the period. Combined: 25,894 GA4 purchases against 36,190 orders, 72%. No store is named.
Picture a Monday morning
GA4 says Paid Social brought you 40 sales last week. Shopify shows 61 orders from people who came from your Meta ads. Somewhere, 21 orders went missing.
So you do what every guide tells you: you open the attribution settings and switch from data-driven to last click, hoping the missing orders turn up under the right channel.
They do not. The total is still 40. Some sales move from Organic to Paid Social and back, but nothing new appears. GA4 can only share out the purchases it saw. The 21 it never saw are not in any model.
The numbers are made up, but the gap is not: in our data, GA4 recorded about three out of every four real orders.
The three models GA4 has today
| Model | Who gets the credit |
|---|---|
| Data-driven (default) | Shared between the clicks before the purchase, weighted by how much each one seemed to matter, based on your property's data. |
| Paid and organic last click | All of it to the last channel the buyer clicked through, ignoring direct visits. Also called last non-direct click. |
| Google paid channels last click | All of it to the last Google Ads click. With no Google Ads click in the path, it falls back to the model above. |
First click, linear, time decay and position-based were removed in November 2023, and nothing has come back since (Google Analytics Help, models explained). Many older guides still list them.
Where to change it, and what happens when you do
- Where: Admin, then Data display, Events, Attribution settings. You need the Marketer role or higher.
- The model rewrites history. Google says a model change "applies to historical and future data". Last month's report will look different tomorrow.
- The lookback window does not. It only applies from the day you change it.
- Lookback defaults: 90 days for purchases (30 or 60 also possible), 30 days for first visits (or 7).
Source: Select attribution settings, Google Analytics Help.
Write the date down before you change the model. Otherwise you will compare months that were counted two different ways without knowing it.
The model only changes some reports
This is the part that confuses most people. The same purchase can sit under three different channels in three reports, by design:
- User acquisition shows where each person first came from. Fixed, the model does not touch it.
- Traffic acquisition shows where each visit came from, by last non-direct click. Also fixed.
- Advertising reports and explorations that show purchases or revenue by source, medium or campaign follow the model you chose.
Source: Scopes of traffic-source dimensions, Google Analytics Help. To see the effect on your own data, open Advertising, Attribution, Model comparison (Google Analytics Help).
So should you change it?
- Keep data-driven if you have steady sales every week and mostly want to see which channels start and help sales, not only finish them.
- Pick Paid and organic last click if you need numbers you can explain to a client or a boss in one sentence, or if you have few sales a month. Google does not publish a minimum for data-driven, and with little data the split moves around a lot.
- Pick Google paid channels last click only if you want GA4 to sit closer to Google Ads, for example to compare the two. It gives Google every sale it touched last.
The bigger problem: the orders GA4 never sees
In our data, GA4 recorded a median 76% of the stores' real orders. In 9 of 16 stores it was under 80%. In none was it above 110%. A model setting moves credit between the purchases GA4 saw. It cannot add the ones it missed.
The usual reasons a purchase never reaches GA4:
- the buyer declined the consent banner, or blocks trackers
- the checkout or payment step runs on another domain and the purchase event is lost on the way back
- the buyer saw the ad on a phone and bought on a laptop, and GA4 cannot join the two without your own login data
Why GA4, Meta and Google Ads never agree
| Counts views without a click? | Default window | Sale is dated | |
|---|---|---|---|
| GA4 | No (only YouTube engaged views) | 90 days for purchases | When the purchase happened |
| Meta Ads Manager | Yes, within 1 day | 7-day click or 1-day view | By Meta's own rules per setting |
| Google Ads | Only its own video views | 30-day click | On the click date |
Sources: Meta Business Help Center, Google Ads Help.
Each tool answers a different question. Use GA4 to see how people move through your site. Use your Shopify or WooCommerce orders as the truth for what sold.
Who does not need to worry about this: if GA4 records more than 90% of your orders and you run one paid channel, the model choice will barely change your decisions.
FAQ
Does changing the attribution model in GA4 change my Google Ads bidding?
Not directly. Conversions sent to Google Ads now have their own attribution settings under Advertising, Conversion management, separate from the reporting model for key events (Google Analytics Help).
Why does GA4 show more direct traffic than I expect?
Visits with no known source land in Direct, for example clicks from some apps and AI chats. In the attribution models, direct visits only get credit when the whole path was direct.
How do I check how many of my orders GA4 misses?
Take a closed month. Compare GA4's purchase count with the number of orders in Shopify or WooCommerce for the same dates and time zone. Under 80% means it is worth fixing the tracking before you argue about models.
We make TrueROAS, which starts from every Shopify and WooCommerce order and shows the ad and visits behind it, so the missing quarter is not missing.
Written by Rasmus Arvidsson, founder. Store numbers combined across 16 stores; no store is named.

