Paid Media Needs an Operating System

Every account I inherit looks busy. Dozens of campaigns, hundreds of ad groups, a naming convention that made sense to someone in 2023. Busy is easy. What is almost always missing is a set of rules that decides what happens next.

That set of rules is the operating system. Without it, an account is a pile of decisions made by different people on different Tuesdays, none of which can be reconstructed six months later.

A pile versus a system

A pile of campaigns produces results. It just cannot tell you why. When performance improves, nobody can name the cause, so nobody can repeat it. When it degrades, the response is to add more campaigns, which makes the pile larger and the signal worse.

A system produces the same results and a reason. The reason is what compounds. It is the difference between an account that is 20% better every quarter and one that is 20% better once and then drifts.

A pile of campaigns produces results. It just cannot tell you why.

What the operating system actually contains

Four things, and they fit on one page. If they do not fit on one page, they will not survive contact with a busy quarter.

• 01A measurement contract. One definition of a conversion, one source of truth, and a written note of what the platform numbers will overstate and by roughly how much. Agreed before the spend, not after the argument.
• 02A testing cadence. How many tests run at once, how long each runs, and the threshold at which you call it. Written down, because the threshold you choose after seeing the data is not a threshold.
• 03A budget rule. What triggers an increase, what triggers a cut, and who is allowed to make the call without a meeting. Most accounts have no rule here, which is why budget moves on sentiment.
• 04A kill list. The things you have already tested and will not test again this year, with the date and the result. This is the most valuable page and the one nobody keeps.

Why the kill list matters most

Teams re-run the same failed test roughly every eighteen months, because the person who ran it the first time has moved on and the result lived in their head. I have watched a company test the same broad-match expansion three times in four years. It lost money on all three occasions.

The kill list turns a team's experience into the account's experience. It is also the cheapest thing on this list to maintain — a shared document and ten minutes at the end of a test.

The compounding part

None of this makes any single quarter dramatically better. That is the honest version. What it does is stop each quarter starting from zero.

An account with an operating system carries its learning forward. Two years in, it is running tests the competition has not thought to run yet, because it stopped re-running the obvious ones a long time ago. That gap is very hard to close with budget.

Start with the measurement contract. Everything else is guesswork until the numbers mean the same thing to everyone reading them.