One variable

From the platform side of the desk you see a great many small advertisers at once, and after a while you stop seeing accounts and start seeing habits.

Two habits, mostly.

The first kind of advertiser has a bad week and changes one thing. New creative, same audience, same budget, same landing page. Then they wait, which is the hard part, and at the end of the week they know something. Maybe the creative was the problem. Maybe it was not. Either way, the list of suspects is one shorter.

The second kind has a bad week and changes everything. New creative, new audience, double the budget, a different offer, a new landing page, all on the same Sunday night. The next week is better, or worse, and they have no idea why. So they change everything again.

The second kind is not lazy. Usually they are working harder than the first. Every single change is a reasonable idea. It is the combination that destroys the information, because five changes produce one result, and one result cannot be shared five ways.

The advertisers who grew, over months, were almost always the first kind. Not because they were smarter. Because they had accumulated a few dozen small, true facts about their own business, and the second kind had accumulated a few dozen weeks of noise.

I still catch myself doing it. A number goes the wrong way and the instinct is to fix all of it, tonight. The discipline is to pick the one thing you most suspect, change only that, and sit on your hands.

A test with five variables is not a test. It is a mood.