Last Click Gives One Channel the Sale and Lies About the Rest
Last-click attribution hands the whole sale to the final touch and hides the channels that did the persuading. GA4's data-driven model splits the credit, but only if the store stops reporting last click as the truth.
A shopper discovers a store through a creator video, comes back twice through organic search, reads three blog posts, and finally buys after clicking a retargeting ad. Last-click attribution gives the retargeting ad 100 percent of the sale and the creator video nothing. The next budget meeting cuts the creator budget, because the report says it never sells anything, and the retargeting budget grows, because the report says it sells everything.
This is not a measurement quirk. It is a budget decision made on a false premise, repeated every month, until the channels that fill the funnel are starved and the channel that closes the sale has nothing left to close. The fix is not a new tool. It is a different model, and a willingness to report the old one as the partial view it is.
What last click hides
Last click is a rule, not a measurement. It assigns all the credit to the final touchpoint, regardless of what happened before it, which means any channel that introduces or persuades is invisible by construction. Paid search and retargeting look like heroes, because they sit at the end of the journey, and prospecting, content, and creators look like waste, because they sit at the start.
The distortion is worst for considered purchases. A shopper who takes nine days and six visits to buy a 200 dollar item will have most of her touches erased by last click, and the report will claim the sale came from a single ad she clicked on the last day. The store then optimizes for the last day, which is the cheapest part of the journey to buy and the hardest part to grow.
What data-driven attribution does differently
GA4's default data-driven model compares the actual path to conversion against the paths that did not convert, and distributes credit across the touches that actually moved the needle. A creator video that appears on converting paths far more often than on non-converting paths gets credit, even if it never closes a sale by itself.
The model needs volume to work. A store with a handful of conversions a month will get noisy, unstable credit splits, and the numbers will jump around from week to week. That is not a reason to keep last click. It is a reason to read the model at a monthly level, and to pair it with a holdout test before cutting a channel's budget.
How to report it honestly
Report both models side by side, and label them. Last click is a record of what closed the sale. Data-driven is an estimate of what contributed to it. Neither is the truth, and a report that shows only one of them is making a claim it cannot support.
Before cutting a channel, run a holdout. Turn the channel off in one region, keep it on in another, and measure the difference in total revenue, not in attributed revenue. If total revenue does not fall when the channel goes dark, the channel was not contributing. If it does fall, the attribution model was right and the budget cut was wrong.
Related reading: Attribution Playbook, Answers, Glossary.
Frequently asked questions
Why does last-click attribution favor retargeting?
Because last click gives all the credit to the final touch, and retargeting sits at the end of the journey. Channels that introduce or persuade, like creators and content, are invisible by construction, even when they did most of the work.
What is GA4 data-driven attribution?
A model that compares paths that converted against paths that did not, and distributes credit across the touches that actually moved the needle. A creator video that appears far more often on converting paths gets credit even if it never closes a sale alone.
Does data-driven attribution work for small stores?
It needs volume. A store with a handful of conversions a month will get noisy, unstable splits. Read the model monthly rather than weekly, and pair it with a holdout test before cutting a channel's budget.
How do I test whether a channel really contributes?
Run a geo holdout: turn the channel off in one region, keep it on in another, and measure the difference in total revenue, not attributed revenue. If total revenue does not fall, the channel was not contributing.