Attribution is the rule an ad or analytics tool uses to assign a conversion and its value to the channels and clicks on a customer's path. It is set by an attribution model and an attribution window. That is why the same order can look different in Google Ads, GA4 and Meta.
Attribution is the rule an ad or analytics tool uses to assign a conversion and its value to the channels and clicks on a customer's path. It is set by an attribution model and an attribution window. That is why the same order can look different in Google Ads, GA4 and Meta.
Customers rarely buy on the first click. They see an ad on Instagram, click a product in Google Shopping a week later, open a newsletter and finally type the store's name into Google. One order, four touchpoints. Attribution decides how much of the order value each one gets.
Each tool only sees its own part of the path. Google Ads does not see the Meta ad, and Meta does not see the Google click. So each one claims the order it took part in, and the revenue claimed by all platforms together usually adds up to more than the store actually took in.
Attribution has two settings. The attribution window sets how long after a click or view a conversion still counts. The attribution model sets how the value is split between the touchpoints inside that window. Example: a 100 EUR order and four clicks — Meta, Performance Max, newsletter, brand search.
| Last click | 100% to the last click | Brand search 100 EUR, others 0 |
| First click | 100% to the first click | Meta ad 100 EUR, others 0 |
| Linear | value ÷ number of touchpoints | 4 × 25 EUR |
| Position-based | 40% first, 40% last, 20% the rest | 40 + 10 + 10 + 40 EUR |
| Data-driven | share based on account data | Google calculates the split from paths that did and did not convert |
The first four are rule-based models. Google has removed them from both Google Ads and GA4, leaving only data-driven and last click. They are still the best way to see how much the result depends on the rule you pick.
Each tool ships with different defaults. The table shows the state as of October 2026.
| Tool | Models | Default window | Note |
|---|---|---|---|
| Google Ads | data-driven (default), last click | 30 days after a click (1–90 possible), 1 day after a view | Records the conversion on the click date |
| GA4 | data-driven (default), paid and organic last click, Google paid channels last click | 90 days for key events (30 or 60 possible), 30 days for acquisition events (7 possible) | Records the conversion on the conversion date, sees organic and other unpaid channels |
| Meta | its own crediting within the window, optional incremental attribution | 7 days after a click, 1 day after a view | Since March 2026 only link clicks count as clicks |
Meta removed the longer view-through windows (7 and 28 days) in January 2026. A gap of tens of percent between tools is normal. What should worry you is when the platforms together report more revenue than the store made.
We dealt with an extreme case at Papírnictví VojTech: campaigns counted add-to-cart as a purchase and also had a duplicate purchase tag. In April 2026 the ads claimed 167% of the store's actual revenue. After the tracking fix on 11 May 2026 it was 54–60% (June–August, Google Ads vs store back office). That is no longer an attribution problem but a tracking one. A model only splits what you send it.
One order, four clicks: a Meta ad, Performance Max in Google, a newsletter link and finally a search for the store's name. Switch the model and watch what each channel gets. Below it you will see what each system claims for itself.
In Google Ads, keep data-driven if it is available to you. Last click gives all the credit to the campaigns that catch people right before they buy, typically brand search. The campaigns that created the demand then look worse than they are, and automated bidding follows that picture.
For splitting budget between channels, do not trust any model inside a single platform. Our rule: the store back office decides (revenue, margin), not the sum of what the platforms claim. Once a month, put revenue, Google cost, Meta cost and margin after ad spend into one table.
At MyDeko, Meta's own reporting shows a ROAS of 1.26 (March–September 2026). Anyone looking only at Ads Manager would undervalue the channel: there, Meta creates the demand and Google catches it. Attribution will not tell you what the ads really added. A test of incrementality will.
Attribution sits in the middle: a conversion defines what counts, attribution decides who gets credit, and incrementality checks whether the ads really caused it.
| Term | Question it answers | Where the number comes from | When to use it |
|---|---|---|---|
| Attribution | Who gets credit for a conversion | Model and window in the tool | Running campaigns inside a platform |
| Conversion | What counts as success at all | Tracking setup | The base attribution cannot work without |
| Incrementality | Would the conversion happen without ads | Test with a control group | Budget and pausing a channel |
| ROAS | Revenue from attributed conversions per unit of spend | Attribution output ÷ cost | Check against break-even |
| CPA | Cost per attributed conversion | Cost ÷ attributed conversions | Campaigns for orders and leads |
Terms that change when attribution changes.
Google Ads, Meta and your email tool will all claim the same order. The sum is higher than reality. Check it against revenue and orders in your store back office.
A new model does not add a single order; it only moves credit between campaigns. It does change the data automated bidding learns from, though. Expect a period of readjustment and avoid other big changes during it.
Google Ads records a conversion on the click date, GA4 on the conversion date. For products with a longer decision cycle the last few days look weak in Google Ads and fill in later. Compare closed periods, ideally full months.
The rule that assigns a conversion and its value to the channels and clicks a customer came through. It is defined by an attribution model (how the value is split) and an attribution window (how long after a click or view a conversion still counts).
Data-driven by default. You can also choose paid and organic last click or Google paid channels last click. Google has removed rule-based models such as first click and linear.
Each tool uses a different model, window and reporting date: Google Ads uses the click date, GA4 the conversion date. GA4 also shares credit with organic search, email and other channels Google Ads cannot see. A gap of tens of percent is common.
Seven days after a click and one day after a view. Since March 2026 only link clicks count as clicks; likes, comments or shares fall into a one-day engagement window. Meta removed the 7-day and 28-day view windows in January 2026.
For running Google Ads campaigns, usually yes, because it also credits campaigns earlier in the path. But it still only splits the conversions Google saw. Whether the order would have happened without ads, only an incrementality test can measure.
Send this and you get a 90-day strategic plan for raising the performance of your marketing. Free, within 48 hours of account access.
Jan gets back to you within 48 hours, usually with two questions and a time. Then the 90-day plan starts.