Incrementality
Don’t trust the model.Make it prove itself.
We turn a channel off in a handful of your markets and leave the rest alone. Whatever gap opens between them is what that channel was really worth. Not attributed. Not modelled. Observed.
Test the channel you argue about mostWhat changes for you.
- You think TV is working.You know what TV is worth, in dollars.
- Every platform claims the same sale.Every channel has one number nobody argues with.
- “Trust the model.”You watched it happen in your own markets.
- Budget fights end with whoever is loudest.Budget fights end.
A number, and how sure we are of it.
Most reporting hands you a number with no range. That is how a test that measured nothing gets sold to you as a win. The range is the honest part.
Result
+18% lift · 95% interval 11% to 25%
The bar is the range the true effect probably sits in. If it touches zero, you did not measure a result — you measured your own uncertainty.
The whole interval is above zero. This channel is doing real work — fund it further and re-test at the higher level.
One test makes every other number sharper.
A test answers one channel at one spend level. Feed it back into the model and it stops guessing where you now have evidence — and gets sharper everywhere near it.
A model with nothing to check it against. The line looks confident. The shaded area is how wrong it could be — and at high spend, it could be very wrong. This is what you are moving budget on when nobody tests.
The questions everyone asks first.
Will I lose sales in the off markets?
A little. That is the price of finding out. We size the test so the loss is small and the answer is readable — and if the channel turns out to be worth it, you turn it back on knowing that for the first time.
How long does it take?
Four to six weeks for most businesses. We fix the end date before we start, and we do not stop early because the numbers look good.
What if I have no regional data?
Then we use a different causal method around a single change instead. You would hear that in week one, not month three.
Can I test everything?
No. Each test costs real spend and real weeks. That is exactly why the model matters — it tells you which test is worth running next.
What if the test says my channel does not work?
Then you just found the most valuable thing on this page. Check the range first though: a test that was too small finds nothing, and that gets misreported as failure constantly.
Do I need a model first?
No. A test is the cheapest thing we do and the easiest to defend. Plenty of clients start with one and decide about a model after.
Pick the channel you argue about most. We’ll settle it.
One test. Four to six weeks. One number nobody in the room can argue with. You do not need a model to start.
Book the two-week audit