What a Marketing Attribution Model Actually Does
A DTC brand hired me in March because their Meta dashboard said Facebook drove 620 conversions last month, Google Ads claimed 480, and GA4 reported 370 total conversions across all channels. Three platforms, three different stories, zero confidence in any of them. The culprit was not a tagging bug (although they had those too). The real problem was that each platform was running a different attribution model with different rules, and nobody on the team understood what those rules were.
A marketing attribution model is a set of rules that decides which touchpoints get credit for a conversion. That is it. No magic, no AI pixie dust, just a framework for dividing up credit. The problem is that every platform picks the framework that makes itself look best, and most marketing teams never question it.
The Models You Will Actually Encounter
Before 2023, Google Analytics offered six attribution models. Then Google removed four of them -- first-click, linear, time-decay, and position-based -- citing that fewer than 3% of conversions in Google Ads used them. Today, your choices in GA4 are data-driven attribution and last click. Meta runs its own model inside Ads Manager. And your CRM probably credits the last UTM it captured.
Here is what each attribution model in marketing actually does.
Last-Click Attribution
Every dollar of credit goes to the final touchpoint before conversion. If a prospect saw your LinkedIn ad, clicked a Google ad, read two blog posts, and then converted from an email, the email gets 100% of the credit.
Why it persists: It is dead simple. CRMs default to it because they capture the last UTM parameters on a form submission. GA4 offers "last click" as an alternative to data-driven attribution.
Where it lies: It systematically undervalues awareness and consideration channels. If you only run last-click, you will eventually cut every top-of-funnel program that feeds your pipeline and wonder why conversions dried up three months later. I have seen this play out more than once with B2B teams -- I wrote about the pattern in Marketing Measurement for Long B2B Sales Cycles.
First-Click Attribution
All credit goes to the first touchpoint. The ad that originally introduced the prospect to your brand gets 100%.
Why it mattered: It helped brand teams justify awareness spend. Some marketers still track it manually in spreadsheets or BI tools.
Where it lies: It ignores everything that happened between discovery and conversion. Google removed it from GA4 in September 2023, and I have not missed it.
Linear Attribution
Credit is split evenly across every touchpoint. Five touches, each gets 20%.
Where it lies: A display impression that someone ignored and the demo request that closed the deal are treated as equally important. Nobody actually believes that.
Position-Based (U-Shaped) Attribution
40% to the first touch, 40% to the last, 20% split across the middle. This was popular with B2B teams because it rewarded both discovery and closing.
Where it lies: The 40/40/20 split is arbitrary. Why not 30/30/40? Nobody knows. It felt reasonable, but it was never empirical.
Time-Decay Attribution
More credit goes to touchpoints closer to conversion. A click one day before converting is weighted more heavily than one 30 days prior.
Where it lies: It penalizes long consideration cycles. For B2B with 90-day sales cycles, this model starves upper-funnel channels that did most of the persuasion work.
Data-Driven Attribution (DDA)
Google's machine-learning model analyzes both converting and non-converting paths to assign credit based on each channel's observed contribution. It is now the default and recommended model in GA4.
Where it lies: DDA is a black box. You cannot see the weights. It requires sufficient conversion volume to function reliably -- Google's documentation notes that properties with low traffic fall back to a modeled last-click approach. And it only sees what GA4 can measure, which means it is blind to offline touchpoints, dark social, and anything lost to consent restrictions.
Here is a summary of how each model distributes credit for the same five-touch journey:
| Model | Touch 1 (Ad) | Touch 2 (Organic) | Touch 3 (Email) | Touch 4 (Blog) | Touch 5 (Direct) |
|---|---|---|---|---|---|
| Last-click | 0% | 0% | 0% | 0% | 100% |
| First-click | 100% | 0% | 0% | 0% | 0% |
| Linear | 20% | 20% | 20% | 20% | 20% |
| Position-based | 40% | 6.7% | 6.7% | 6.7% | 40% |
| Time-decay | ~5% | ~10% | ~15% | ~25% | ~45% |
| Data-driven | Varies | Varies | Varies | Varies | Varies |
That marketing attribution model example should make one thing obvious: the "best" channel changes depending on which row you read. Same data, six different budget decisions.
Why Every Platform Tells a Different Story
Each ad platform uses an attribution model in digital marketing that is scoped to its own walled garden and tuned to make that platform look good.
Google Ads defaults to data-driven attribution within its own ecosystem and counts conversions within a 30-day click, 1-day view window by default.
Meta Ads defaults to 7-day click, 1-day view attribution. In January 2026, Meta removed the 7-day and 28-day view windows entirely. And after Apple's App Tracking Transparency rollout -- with global opt-in rates still around 38% as of Q1 2026 -- Meta increasingly relies on modeled conversions that you cannot independently verify.
GA4 sees cross-channel behavior but is constrained by consent loss. Safari's Intelligent Tracking Prevention caps client-side cookies at 7 days (24 hours when a user arrives via a tracking parameter), which means GA4 cannot stitch sessions for returning Safari users after a week. I covered the practical fixes in First-Party Data Strategy After the Cookie U-Turn.
Your CRM typically credits the last known UTM values on the lead record. That is last-touch by accident, not by design.
The result: Google says Google drove the conversion. Meta says Meta drove it. Your CRM says email did. Everyone is partially right and collectively useless for budget decisions.
What to Actually Trust
After years of building attribution stacks for clients, here is what I recommend.
Stop looking for the one true model
No single marketing attribution model will give you the full picture. Every model has structural blind spots. The question is not "which model is correct?" but "what decisions am I trying to make, and which combination of views reduces my risk of a bad one?"
Use data-driven attribution as your baseline, not your answer
DDA in GA4 is the least-bad default for digital channel allocation. It at least attempts to weight touchpoints by their observed impact rather than by arbitrary rules. But treat it as one input, not gospel. Export the raw path data from BigQuery and run your own analysis when the stakes are high -- I walk through the setup in GA4 BigQuery Export Setup and First Attribution Queries.
Layer in incrementality testing
The gold standard for attribution is not a model at all -- it is experimentation. Geo-holdout tests, matched-market tests, or simple on/off tests where you pause a channel in a subset of markets and measure the lift. This tells you what a channel actually caused, not just what it touched.
Build a unified view across platforms
Stop comparing Google Ads conversions to GA4 conversions to CRM conversions as if they should match. They will not match because each system counts differently. Instead, build a single source of truth -- usually in your data warehouse -- that deduplicates conversions and maps each one to the touchpoints you can verify. If the gap between platforms and reality is bigger than you expected, that is exactly the kind of problem I help teams diagnose.
Triangulate, do not optimize to one metric
The most trustworthy marketing measurement combines at least three lenses: multi-touch attribution for tactical channel allocation, incrementality tests for strategic budget shifts, and blended efficiency metrics (like blended CAC or MER) for the boardroom. If all three point in the same direction, act with confidence. If they diverge, investigate before spending.
Common Traps I See Repeatedly
Trusting view-through conversions at face value. A "view" in Meta's attribution model means the ad loaded in the viewport. The user may never have consciously noticed it. If you report view-through conversions alongside click-through conversions without separating them, you are inflating your numbers.
Ignoring consent-mode gaps. If you are running Google Consent Mode v2 in the EEA, Google is modeling a portion of your conversions. The modeled data is directionally useful but it is an estimate, not a measurement. Your attribution model marketing reports should flag what percentage of conversions are modeled.
Comparing attribution windows across platforms. Google Ads default 30-day click window versus Meta's 7-day click window means Meta is structurally undercounted relative to Google for any conversion that takes more than a week from first click. Normalize windows before comparing.
Running DDA on thin data. Google does not publish a hard minimum, but in my experience -- and that of most practitioners I talk to -- if your GA4 property has fewer than roughly 300 conversions per month on a given key event, data-driven attribution tends not to have enough signal to model reliably. In these cases, last-click may actually give you more stable -- if less sophisticated -- reporting.
FAQ
What is a marketing attribution model?
A marketing attribution model is a set of rules or algorithms that determines how credit for a conversion is distributed across the marketing touchpoints a customer interacted with before converting. Different models assign credit differently, which directly affects which channels appear most valuable in your reports.
Which attribution model does GA4 use by default?
GA4 uses data-driven attribution as its default model. This machine-learning approach analyzes converting and non-converting paths to assign credit based on observed impact. GA4 also offers last-click as an alternative, but the four rule-based models (first-click, linear, time-decay, and position-based) were removed in September 2023.
Why do Google Ads and GA4 show different conversion numbers?
Google Ads counts conversions using its own attribution model scoped to ad interactions, while GA4 evaluates all traffic sources together. They also differ on attribution windows, counting methodology, and how they handle consent gaps. These structural differences mean the numbers will rarely match exactly.
Can I still use first-click or linear attribution?
Not natively in GA4 or Google Ads, as Google removed both models in 2023. You can still calculate them manually using raw event data exported to BigQuery, or use third-party tools that support rule-based models. However, most practitioners find data-driven attribution more useful than reverting to static rules.
How many conversions does GA4 need for data-driven attribution to work?
Google does not publish a hard minimum, but properties with low conversion volume will see GA4 fall back to a modeled last-click approach. In practice, most consultants observe that data-driven attribution produces stable, differentiated results once a key event reaches roughly 300 or more conversions per month.
Not sure which attribution model to trust -- or whether your tracking is even capturing the full picture? Let me take a look. I will tell you exactly what is broken and what to fix first.