Open any marketing dashboard and one channel will look like the hero. It usually has the lowest cost per sale and the most conversions, and it is tempting to put every spare dollar behind it. In many accounts that channel is not the hero. It is just the last place a customer clicked before they bought.
That is not the same as the place that made them want to buy. A good report keeps the two apart, and a lot of reports do not.
Most tools give all the credit to the final touch. A person sees a video on Monday, reads a newsletter on Wednesday, searches your brand name on Friday and buys. The report says search made the sale. Video and email did the work of making them look for you, and neither gets a single point.
Nobody is lying here. The tool is doing exactly what it was set up to do. The problem is that the setting is the easiest one to leave alone, so it never gets questioned. Over a few months the budget drifts toward the channels that collect credit and away from the ones that create demand. Then the demand dries up, and the report cannot explain why.
We see this most with branded search. It looks cheap because those people were already on their way. Pay for more of it and you mostly buy clicks you would have received for free.
The last click is a receipt, not a reason.
You do not need a data scientist to spot it. Check your own account for these three patterns.
Branded search grows faster than everything else. If people search your name more each month, something else is making them aware of you. Find out what.
Pausing a quiet channel hurts search a few weeks later. The drop comes late, so it is easy to blame the wrong thing. Mark the dates when you make changes.
Your best channel stops responding to extra budget. If doubling the spend gives you a few more sales instead of double, it was only ever catching people who were coming anyway.
One of these is a warning. Two is a pattern. Three means the report has been telling you a comfortable story.
You can find the truth with a plain holdout test. It takes four weeks and costs very little.
Pick one channel you suspect is over-credited, and one region or audience where you can switch it off.
Switch it off there for four weeks. Leave every other region exactly as it is.
Write down the start and end dates, and what result would make you stop early.
At the end, compare total sales in the two groups. Ignore the channel reports completely.
The gap between the two groups is what the channel was really adding. If sales fell by a lot, the channel matters and the report was right. If they barely moved, you have found money you can spend somewhere that does more.
Keep the test honest. Change one thing at a time, do not run it over a holiday, and do not peek at week two and call it. A short, clean test beats a long, messy one.
Imagine a small online shop that sells outdoor gear. Its dashboard says search brings 60 percent of sales at the lowest cost, so the owner plans to double the search budget.
Before doing it, we run a holdout. For four weeks, we pause the paid brand search in one of five regions. At the end, total sales in that region are down 3 percent, while the other four are flat. If search had really produced 60 percent of the sales, that region would have dropped by far more.
The shop learns that most of those searchers would have arrived anyway. The doubled budget would have bought very little. Instead, the owner moves a slice into video and email, the channels that were quietly creating the searches, and keeps testing.
Your numbers will be different. The method is the same: switch something off in one place, compare total sales, and believe the sales over the report.
Do not move the whole budget on one test. Shift a small slice, around ten percent, toward the channel that created demand, and watch total sales for another four weeks. If they hold or grow, move another slice. If they dip, you have learned where the floor is, and you can put the money back.
Then change the report so it cannot fool you again. Show total sales next to the channel numbers, and show each channel next to what it cost. The number should always sit beside its sum. That is how we write ours, and every figure is checked by a second person before it goes out.
A channel that gets all the credit is not always wrong. But you should know why you believe it, and you should be able to show the sum.
Pass it on
If you only have a minute, these are the three things to keep.
