What are attribution models?
A customer rarely experiences your marketing through one clean path.
They might discover a product through organic search, see a Meta ad two days later, return through Google Shopping, open an email, and finally type your brand name into Google before purchasing.
Which channel gets credit for the sale?
That is the question attribution tries to answer.
An attribution model is a rule or algorithm that determines how conversion credit is distributed across the marketing interactions recorded before a conversion.
The important word is credit.
Attribution does not tell us with certainty which interaction caused the purchase.
It tells us how a measurement system has decided to assign the purchase based on the data it can observe and the attribution rules being used.
Attribution is not the same as causation
This distinction is one of the most important parts of modern marketing measurement.
Imagine this customer journey:
Instagram ad → organic search → email → Google Search ad → purchase
A last-click model could assign the entire sale to Google Ads.
A first-click model could assign it entirely to Instagram.
A linear model could split the credit equally.
A data-driven system might divide credit differently again.
The order never changed.
Only the measurement model changed.
Attribution changes who receives credit. It does not change what actually happened.
This is why attribution reports should support marketing decisions rather than be treated as an unquestionable version of reality.
Why attribution is still useful
If attribution is imperfect, why use it?
Because marketing teams still need a consistent way to analyse complicated customer journeys.
Used carefully, attribution can help answer questions such as:
- Which channels appear early in converting journeys?
- Which channels often appear close to purchase?
- How does conversion credit change under different models?
- Which campaigns contribute across longer paths?
- Are we undervaluing channels because we only look at the final click?
- How are automated advertising systems assigning conversion value?
It gives the team a framework for analysing journeys that would otherwise be reduced to “the final click won.”
Three things determine an attribution result
Before comparing models, it helps to understand that the attribution model is only one part of the measurement system.
1. What the platform can observe
The system needs to connect marketing interactions with a later conversion.
That becomes harder when:
- Customers switch devices
- Cookies are unavailable
- Consent is not granted
- Ad blockers remove measurement
- Customers use different browsers
- Offline interactions are involved
Some platforms use first-party data, consented signals, and modelling to fill part of those gaps.
But no marketing platform observes every customer interaction perfectly.
2. The attribution model
The model decides how available credit is distributed among the interactions that were recorded.
3. The attribution window
The window determines how far back the system looks for an eligible interaction.
A seven-day click window and a 30-day click window can produce different attributed results even when the model itself remains unchanged.
This distinction matters:
Model = how credit is divided.
Window = which interactions are eligible to receive credit.
The classic attribution models
Several attribution models have shaped how marketers think about customer journeys.
Some of them are no longer available as selectable models in major Google advertising and analytics products, but they remain useful for understanding the logic behind attribution.
1. Last-click attribution
Last-click gives all conversion credit to the final eligible marketing interaction before the conversion.
For example:
Organic search → Meta ad → email → Google Ads → purchase
Under a simple last-click model:
Google Ads receives 100% of the credit.
Why it is useful
Last-click is easy to understand.
It clearly shows which interaction immediately preceded the purchase.
What it misses
It can undervalue the interactions that created awareness, helped the customer compare options, or brought them back earlier in the journey.
A channel may be great at introducing new customers but rarely receives last-click credit.
2. First-click attribution
First-click does the opposite.
It assigns all conversion credit to the first recorded marketing interaction.
For the same path:
Organic search → Meta ad → email → Google Ads → purchase
organic search would receive 100% of the credit.
Why marketers used it
It emphasises customer acquisition and discovery.
That can help teams think about which channels introduce people to the brand.
What it misses
It ignores everything that happened afterwards.
The channel that created initial awareness may not have been enough to produce the final purchase.
Important: first-click attribution is now mainly useful as a conceptual model. Google removed first-click as a selectable attribution model from GA4 and Google Ads.
3. Linear attribution
Linear attribution divides conversion credit equally across the recorded interactions.
If the customer touched four channels:
Organic → Meta → email → Google Ads → purchase
each interaction would receive 25%.
Why it can be useful conceptually
It acknowledges that several touchpoints may contribute to a purchase instead of declaring one interaction responsible for everything.
The limitation
There is no reason to assume every touchpoint contributed equally.
A casual social interaction and a high-intent product search may have very different roles.
Linear attribution is also no longer available as a selectable model in GA4 or Google Ads.
4. Time-decay attribution
Time-decay gives progressively more credit to interactions that occurred closer to the conversion.
For example:
Instagram → organic search → remarketing → email → purchase
email might receive the most credit, remarketing somewhat less, and Instagram the least.
The reasoning
The model assumes interactions closer to the final purchase are more influential.
That can seem intuitive for short sales cycles or promotional journeys.
The limitation
The assumption itself may be wrong.
An early review article or product-discovery ad may have been the reason the customer considered the brand at all.
Google has also retired time-decay as a selectable attribution model.
5. Position-based attribution
Position-based attribution, sometimes called U-shaped attribution, historically gave greater weight to the beginning and end of the journey.
A common implementation assigned:
- 40% to the first interaction
- 40% to the final interaction
- 20% divided among the interactions in between
The idea was to recognise both discovery and conversion.
But those percentages are still arbitrary.
They reflect a rule chosen in advance, not evidence that those interactions genuinely contributed in those exact proportions.
Google has also removed position-based attribution from Google Ads and GA4.
6. Data-driven attribution
Data-driven attribution takes a different approach.
Instead of starting with a fixed rule such as:
“The final click receives everything.”
or:
“Every interaction receives the same amount.”
the system uses observed conversion data to estimate how different interactions contribute across customer paths.
How Google's data-driven attribution works
Google's systems compare patterns across converting and non-converting journeys and assign fractional credit to interactions based on their estimated contribution.
That means a journey might end up reported as:
- Organic search: 0.24 conversions
- Paid Search: 0.41 conversions
- Email: 0.35 conversions
The customer did not make 0.24 of a purchase.
The decimals represent fractional attribution credit.
Data-driven does not mean perfectly accurate
This is where marketing language sometimes becomes misleading.
Data-driven attribution is more flexible than fixed rules.
But it still works within the data available to the platform.
It can be affected by:
- Missing interactions
- Consent and privacy restrictions
- Cross-device behaviour
- Conversion tracking quality
- The attribution window
- Which channels the platform can observe
So we would not describe data-driven attribution as automatically discovering the “true” contribution of every marketing channel.
It is still a model.
What attribution models can you actually use in GA4 today?
This is where older attribution guides can become confusing.
Google Analytics 4 no longer offers the full list of classic rule-based models.
As of 2026, the Attribution reports support three approaches:
Data-driven attribution
Credit is distributed according to the property's observed data and Google's attribution model.
Paid and organic last click
The final eligible non-direct channel receives the credit.
For example:
Paid Social → Organic Search → Email → Direct → purchase
email can receive the credit because direct traffic is generally ignored when another eligible interaction exists.
Google paid channels last click
This gives conversion credit to the final Google Ads interaction when one exists.
If there is no eligible Google Ads interaction, GA4 falls back to paid and organic last click.
First user, session source, and attribution are different things
This causes a lot of confusion in GA4.
Seeing:
First user source
does not mean you switched to a first-click attribution model.
It answers a different question:
How was this user first acquired?
Likewise:
Session source
describes how the current session was acquired.
The selected attribution model mainly affects event-scoped conversion and revenue reporting.
That means a channel can appear differently depending on whether you are looking at:
- First user acquisition
- Session acquisition
- Attributed key events
- Attributed revenue
Before comparing reports, check which scope you are actually using.
What attribution models does Google Ads use?
Google Ads has also simplified its attribution options.
Today, the main choices are:
- Data-driven attribution
- Last-click attribution
Data-driven attribution is the default for most conversion actions.
Google no longer supports first-click, linear, time-decay, or position-based attribution in Google Ads.
Google Ads attribution has a narrower job
Another important distinction:
Google Ads attribution evaluates interactions within Google's advertising ecosystem.
It should not be interpreted as a complete independent comparison between:
- Google Ads
- Meta
- Organic Search
- Affiliates
GA4 can provide a broader cross-channel view.
Individual advertising platforms primarily optimise and report within the data they can observe.
Why Google Ads and GA4 rarely show exactly the same numbers
This is one of the most common attribution questions we hear.
A store records 500 orders.
GA4 reports one number.
Google Ads reports another.
Meta reports another.
This does not automatically mean something is broken.
Differences can come from:
- Different attribution models
- Different attribution windows
- Different event timestamps
- Different identity methods
- Click versus view-based interactions
- Consent
- Cross-device modelling
- Platform-specific reporting logic
There may also be genuine implementation errors.
That is why attribution discrepancies should be understood, not automatically eliminated.
The store backend is still your commercial source of truth
Attribution platforms answer:
Which marketing interaction should receive credit for this sale?
Your commerce system answers:
Did the sale actually happen?
Those are different jobs.
If the store records €100,000 in completed revenue while marketing platforms collectively claim €160,000, you have not secretly made €160,000.
Multiple systems may be assigning credit to some of the same sales.
For eCommerce reporting, we therefore separate:
- Commercial truth: orders, revenue, refunds, margin
- Attribution: how marketing systems distribute credit for those outcomes
This distinction becomes particularly important when comparing channel ROAS.
Do not add attributed revenue from platforms together
This mistake can make marketing performance look much better than reality.
Imagine one customer:
- Clicks a Meta ad
- Later clicks a Google ad
- Purchases for €200
Under their respective reporting rules, both platforms may report some or all of that €200.
Adding:
Meta attributed revenue + Google attributed revenue
does not necessarily equal total business revenue.
The same transaction may appear in both systems.
Always reconcile marketing reporting against actual orders.
Attribution can influence automated bidding
Attribution is not only a reporting choice anymore.
It can also affect optimisation.
Google Ads automated bidding uses conversion data to decide which auctions and users appear more valuable.
If data-driven attribution gives some conversion credit to an earlier Google Ads interaction, that value can become part of campaign optimisation.
This is one reason changing an attribution model should not be treated as a cosmetic dashboard setting.
It changes how conversion value is reported across eligible interactions and can influence the signals used for bidding.
Bad conversion tracking makes every attribution model worse
No attribution model can rescue inaccurate input data.
If Purchase fires twice, the attribution model distributes credit for two purchases.
If failed payments are recorded as sales, the model attributes revenue that never existed.
If transaction values or currencies are wrong, the model works with those wrong values.
Before debating attribution philosophy, validate:
- Purchase fires only for valid orders
- Transaction IDs are passed correctly
- Revenue values are correct
- Currency is correct
- Refund logic is understood
- Consent is implemented correctly
Measurement quality comes before attribution sophistication.
Which attribution model should you use?
In 2026, this question is simpler than many older guides make it sound.
For GA4 reporting
Data-driven attribution is a useful default when you want conversion credit distributed across eligible paid and organic interactions using the property's data.
Paid and organic last click remains useful when you want a simpler model that is easier to explain.
Google paid channels last click is more specialised and intentionally favours the final Google Ads interaction when one exists.
For Google Ads
Data-driven attribution is Google's default for most conversion actions.
Last click remains available if the business intentionally wants that simpler reporting logic.
The important thing is not to search for the theoretically “perfect” model.
Use a consistent model, understand what it measures, and know how it affects reporting and optimisation.
Use model comparison instead of arguing from intuition
When attribution models produce different channel results, the useful question is not:
Which model makes our favourite channel look best?
Instead, compare them.
For example, Google Ads lets marketers compare data-driven and last-click attribution.
This can reveal campaigns or keywords that receive more credit once earlier interactions are considered.
GA4 attribution reports can likewise help show how credit changes under different reporting approaches.
The difference between models is information.
It shows how dependent your conclusions are on the rule used to assign credit.
Attribution is not incrementality
This is the next step once attribution is understood.
Suppose Google Ads receives credit for 1,000 purchases.
That does not automatically mean those 1,000 purchases would have disappeared if the ads had been turned off.
Some customers may already know the brand.
Some may have purchased through another channel.
Some may genuinely have been persuaded by the advertising.
Attribution asks:
Which observed marketing interactions receive credit for this conversion?
Incrementality asks:
How many additional conversions happened because the marketing activity existed?
Those are different questions.
Experiments can answer causal questions more directly
Where sufficient scale and appropriate tools are available, holdout or lift experiments compare groups that receive marketing activity with groups that do not.
The difference can provide evidence of incremental impact.
This is much closer to asking:
What would have happened without the campaign?
Attribution alone cannot answer that counterfactual.
A practical eCommerce measurement stack
Rather than asking one attribution platform to answer everything, we prefer several measurement layers.
| Layer | Question |
|---|---|
| Store backend | How many real orders, refunds, and euros did we generate? |
| GA4 | How do customers arrive, navigate, and convert across channels? |
| Ad platforms | Which advertising interactions receive conversion credit and optimisation signals? |
| Attribution reports | How does credit move when the attribution logic changes? |
| Experiments / lift studies | Did the marketing activity produce conversions that would not otherwise have happened? |
No individual layer replaces all the others.
Five attribution mistakes we see in eCommerce reporting
1. Treating attributed revenue as unique revenue
Several platforms may take credit for the same order.
2. Comparing platforms without checking attribution windows
Different eligibility windows can make one channel appear stronger even before the model is considered.
3. Believing data-driven means objectively correct
It is still based on available observations and modeling.
4. Changing attribution models whenever results look uncomfortable
A model should create consistency, not provide a way to manufacture a preferred answer.
5. Debating attribution while tracking is broken
Duplicate purchases and wrong revenue values are much larger problems than whether an earlier click deserves 20% or 30% of the credit.
How we approach attribution at Creative Brackets
For eCommerce accounts, attribution starts with measurement quality.
1. Validate the commercial event
We check whether the Purchase event represents a real completed order and carries the correct value, currency, and transaction ID.
2. Understand each reporting system
We document what GA4, Google Ads, Meta, and other relevant tools are actually counting.
3. Review attribution settings and windows
We make sure reports are being interpreted according to their actual attribution logic.
4. Reconcile against the store
Marketing attribution should remain grounded in actual order and revenue data.
5. Use the result to ask better questions
Instead of saying:
“Meta generated 37% of all our revenue.”
we might ask:
“Meta is receiving substantial conversion credit. Does that remain true under a different attribution view, and what happens when spend changes?”
That creates a much more useful marketing discussion.
See more about our eCommerce digital marketing and analytics approach →
Give credit without confusing it with truth
Attribution is useful because customer journeys are complicated.
But the neat percentages in an attribution report can create a false sense of certainty.
A model does not know exactly what was going through a customer's mind.
It works with the interactions it can observe and distributes credit according to a defined system.
That is enough to make attribution valuable.
It just means we should ask the right question.
Not:
“Which channel caused this sale?”
but:
“How is this measurement system assigning credit, what assumptions does it make, and is that view useful for the decision we need to make?”
Once that distinction is clear, attribution becomes much easier to use without letting it run the marketing strategy.
Not sure whether your marketing reports are telling the same story?
Creative Brackets can review your GA4 setup, advertising conversions, attribution settings, and purchase tracking to identify gaps before you use the data to make budget decisions.
FAQ
1. What is an attribution model?
An attribution model is a rule, set of rules, or algorithm that determines how conversion credit is assigned across recorded marketing interactions before a conversion.
2. Which attribution models are available in GA4 in 2026?
GA4 Attribution reports currently support data-driven attribution, paid and organic last click, and Google paid channels last click. Google removed first-click, linear, time-decay, and position-based attribution models in 2023.
3. Which attribution models does Google Ads support?
Google Ads currently supports data-driven and last-click attribution. Data-driven attribution is the default for most conversion actions. The older first-click, linear, time-decay, and position-based options are no longer supported.
4. Is data-driven attribution the most accurate attribution model?
It can use observed account data more flexibly than a fixed rules-based model, but it is not a perfect representation of causality. Its output still depends on the interactions, identity signals, conversion tracking, attribution windows, and modeling available to the platform.
5. Why does GA4 show different revenue from Google Ads or Meta?
The platforms can use different attribution logic, windows, identity methods, timestamps, modeling, and observable interactions. Differences are therefore expected. Large or sudden unexplained differences should still be investigated for tracking problems.
6. Does first user source in GA4 mean first-click attribution?
No. First user source describes how the user was initially acquired. Session source describes the source of a particular session. Attribution models determine how credit for key events and associated revenue is distributed. These are different measurement scopes.
7. Should we add Google Ads and Meta attributed revenue together?
No. Both platforms can claim credit for the same transaction according to their own attribution rules. Compare platform attribution with actual store revenue rather than assuming attributed revenue from different systems represents unique sales.
8. Is attribution the same as incrementality?
No. Attribution decides how recorded conversions receive marketing credit. Incrementality attempts to measure how many additional conversions happened because the marketing activity existed. Controlled experiments and lift studies are better suited to causal incrementality questions.
9. Which attribution model should an eCommerce business use?
There is no universal model that reveals objective truth. Data-driven attribution is a useful default in Google's current ecosystem, while last-click provides a simpler view. The important part is understanding the selected model, keeping measurement consistent, and validating decisions against actual commercial results.
