Is that channel actually better?
This isn’t in the product yet.
30-second version
A channel can look strong because it lands on your best page, on phones, on Fridays. A fair comparison accounts for page, device, and day, then looks at what is left.
Do it
- Raw is not adjusted. The sample chart shows a solid bar and a dashed bar. The gap is the point of the method.
What you'll see
Solid gel bars are raw. Dashed bars are the adjusted comparison. SAMPLE.
Under the hood
A controlled regression includes the channel plus the confounders you can measure. It estimates the remaining association, not magic. It still needs a design if you want the word “caused.”
When not to trust it
- This is not shipped. The live channel list is a count, not this model.