A customer sees a video, clicks a social post, reads a blog article, gets an email, then searches the brand name and buys. Ask three platforms who deserves credit, and you will get three confident answers. The video platform claims the view. The search platform claims the click. The analytics platform credits the last channel before conversion. All three are reporting something real, and none of them is the whole truth. That is the attribution problem.
Last click is fast, simple, and misleading
Google Analytics describes last click plainly: paid and organic last click attributes 100% of the key event value to the last channel the customer clicked through before converting, ignoring direct traffic unless the path is entirely direct. In Google's own example, a path of Display, Social, Paid Search, then Email gives 100% to Email.
That model is easy to explain and easy to optimize against, which is why it survives. It is also systematically unfair to everything that happened earlier. The video that introduced the business gets nothing. The post that kept it familiar gets nothing. The branded search ad that closed a sale it did not create gets everything. Run a budget on last click alone for long enough, and you will cut the awareness work that feeds the closing work, then wonder why the closing work stops converting.
Data-driven attribution is fairer, and still partial
Data-driven attribution, the default model in Google Analytics, works differently. Google describes it as a machine learning model that evaluates both converting and non-converting paths, compares what happened with what could have occurred, and assigns fractional credit based on how much each touchpoint changed the odds of a key event. It does not use a fixed rule. It learns from your account's own paths.
That is a genuine improvement, and it comes with an honest caveat: it can only learn from the journey it can see. It sees sessions and clicks inside the property, within the lookback window. It does not see impressions that never became a session, offline conversations, or touchpoints inside walled gardens it cannot observe. A fairer model of a partial journey is still a partial journey, and a business that treats it as complete will still misread its funnel.
Compare models, then settle the disagreements
The practical move is not to pick one model and defend it. It is to run two views side by side, last click as the baseline and data-driven as the fairer read, and focus on where they disagree. Agreement means the decision is probably safe either way. Disagreement is where budget moves are actually being made, and disagreement is where you need better evidence.
That evidence comes from tests, not from a third dashboard. Hold out a channel or a region, run a conversion lift study, and see what happens to sales when the channel is reduced. Google's own data-driven methodology uses a holdback comparison at its core, contrasting exposed users with similar users who were not exposed. Your budget decisions deserve the same discipline: if a channel's credit disappears the moment you stop looking at its own report, it was never that channel's credit to begin with.
Two guardrails keep this honest. First, data-driven models need enough conversion volume to learn from, and a low volume account will see results that look a lot like last click. Second, no model recovers information the tracking never captured. Clean, consistent campaign labels are the floor everything else stands on, which is why tagging comes before modeling in any serious measurement setup.
How we can help
We build measurement that a business can actually decide from. That starts with clean tracking and consistent campaign labels, so every channel reports in the same language. Then we compare attribution models instead of worshipping one, and we settle the big disagreements with holdout and lift tests before moving budget. If your platforms disagree about who deserves credit, and your budget follows whichever report you opened last, the fix is not a new dashboard. It is a measurement discipline that asks what would have happened anyway.
Takeaway
Last click answers fast, and wrong, by giving everything to the final touch. Data-driven attribution answers fairer, by spreading credit according to contribution, and it still only sees the journey it can observe. Compare the models, trust the disagreements as the signal, and settle big budget moves with a test. Credit is a budget decision, not a reporting cosmetic, so treat it like one.
Which channel in your report is getting credit it did not earn?
Data over opinions.
Frequently asked questions
What is last click attribution?
It gives 100% of the conversion value to the last channel clicked before converting, ignoring direct traffic unless the path is entirely direct. Google Analytics uses it as one of its available models, alongside data-driven attribution.
What is data-driven attribution?
A machine learning model that compares converting paths with non-converting paths and assigns fractional credit based on how much each touchpoint changed the odds of conversion. It is the default model in Google Analytics.
Why do platforms report different numbers for the same sale?
Each platform counts differently and sees a different slice of the journey. One counts views, one counts clicks, one counts sessions, and none sees the full path including offline and walled garden touchpoints.
When should a business use a holdout test?
When two models disagree about a channel that carries real budget. Pausing or reducing the channel in a controlled test shows whether sales actually fall, which is stronger evidence than any attribution report alone.
Sources
Relevant Tools
Want measurement you can actually decide from?
Get one data-backed marketing insight every week. The same no-fluff analysis behind this article, applied to what is actually moving budgets.
Free forever. Unsubscribe anytime.
Connect with JC