Attribution decides who gets the credit for a sale, not whether it happened. Switch the rule and the channel you were ready to cut can look twice as valuable.
Attribution decides who gets the credit for a sale. It doesn't decide whether the sale happened. Run one quarter through four models and the total holds still while the per-channel numbers move all over the place. So before you cut the channel sitting at the bottom of the table, have a look at which rule put it there.
Open your analytics and there's a setting most people never touch. It decides how the credit for a sale gets shared across every channel that touched the customer on the way in. Last click, first click, linear, time-decay, whatever your tool calls them.
Change it, and the table you use to judge your marketing rearranges itself.
Same sales. Same customers. A different story about who earned them.
That's the whole point of this piece. A model is a way to read a decision. Treat it as the final verdict on a channel and you'll get the channel wrong. Let me show you with real numbers.
I work with a brand turning over around $4M in tracked revenue a quarter. I took a single quarter and ran it four times, same period, same channels, changing only the attribution model between runs.
The total held. It came back at that $4M every time. What moved was where the credit sat, and it moved a long way.
On last click, Facebook Ads looked like the softest of the big spenders. Share the credit across the whole journey instead and its number came out close to double. Same spend, same people, same three months.
Google Ads swung by more than a third, depending on whether the first touch or the last touch took the credit. Email shifted by around half on the same test.
Nothing changed in the business. The rule changed.
There's a reason all four numbers hold up at once.
For this brand, the average run from first touch to order is the best part of three months. Only about one sale in five came from a single touch. The rest took two or more, and a long tail ran well into double figures before anyone bought.
Now ask a model to take a journey that long, across a dozen touchpoints, and pin the whole sale on one line. It can't. None of them can.
A customer sees you on Instagram and forgets. Three weeks later an ad jogs the memory. An email lands, they search your name, and the order finally comes through. Which of those made the sale? Every one of them. That's the answer, and no single number will give it to you.
Google Ads didn't bring that customer in cold. Organic social and email had already done the warming, and Google took the last step. Score it as a solo act and you'll pay it wrong in one direction or the other. Score it as one player in a side, and it makes sense.
Attribution still earns its place in all this. It's a lens, and a lens does its job as long as you remember you're looking through one. The pattern I've found works comes down to a few habits.
Pick one primary model and stay with it, so a number means the same thing this week as it did last week. Then, before any real budget call, run the same period through a second model. If a channel looks strong through one and weak through the other, you've found a channel doing work you can't see from a single angle. That's information and not a fault in your analytics.
Keep an eye on the blended figure. Everything you sold over everything you spent to sell it. It tells you whether the whole machine is getting more or less efficient, and it doesn't care which model you favour. When the per-channel numbers pull against each other, blended is your tie-breaker.
Hold your three views loosely. The ad platform's own figure, your analytics and your blended efficiency will never line up, and that's fine. Each was built to measure a different thing. Use the platform number to optimise inside the platform, use your analytics for direction and blended for the truth about the whole.
And when it matters whether a channel earns its place, don't argue the models. Test it. Turn the channel down or off for a set stretch and watch total sales. It answers the question every model dodges: what would we have sold without this channel in the mix?
The question was never who gets the credit. It's what's the smallest change to the mix that lifts the whole store. You're running a team. You don't rank a team by whoever tapped in the final goal.
Attribution is a rule about credit, and changing the rule changes the numbers on the page while the sales stay exactly where they were. One quarter showed a channel going close to double on the model alone. Your biggest line is often Direct, the one you control least, and last click flatters it. Journeys this long mean no single model is ever the full truth.
So lean on attribution to guide decisions and to spot what's shifting over time. Lean on blended numbers and real tests to decide whether a channel is earning its keep.
Here's one thing to do this week. Open your reporting, find the model it's set to, and write it down. Then switch it, run the same quarter again, and see which channels jump.
The ones that move the most are the ones a single view has been getting wrong. Give those a proper look before the next budget call.
