Scaling ads without breaking the learning phase

When a campaign starts working, the temptation is always the same: double it. I understand the instinct — you finally found a winner, you want more of it, today. And that instinct is exactly how people break their own winners.

The learning phase is real

On Meta there's something called the learning phase. Google has its version too. When a campaign launches, the platform spends a stretch of time learning who your ideal customer is — who clicks, who buys, who ignores you. When it gets through that phase, the system has a working picture of the person it should be finding for you.

Double the budget overnight and you can reset that learning. The platform starts figuring out your customer all over again, and the performance you were so excited about goes with it. You didn't scale the winner. You re-entered the lottery.

So here's the instruction I actually give: take the top performers and raise them ten percent. Protect the learning phase. Even a small raise can occasionally trigger a reset, but at that size the risk is low and the compounding is real. Ten percent, again and again, beats doubling once and starting over.

Watch frequency like it owes you money

The other number I monitor closely across every advertising campaign is frequency — how many times the same person has been shown the same ad. Showing someone your ad two or three times is fine. That's normal persuasion. But when frequency starts getting close to four, your numbers start to suffer, because the platform is charging you to repeat yourself to people who have already decided.

High frequency is not a media-buying problem. It's a creative problem. It means the ad is tired, and the fix is a refresh: new angles, new creative, new reasons for the same person to look again — or a wider audience so the platform has somewhere new to go.

Let the machine do the watching

None of this is complicated. What makes it powerful is doing it every day without fail — and that's exactly what most advertisers can't sustain, because checking dashboards is boring and the days are full.

This is where AI earns its seat. With my ad account connected, I ask one question in the morning and get the read: which campaigns are through learning, which ones deserve their ten percent, which ads are pushing a frequency of four and need new creative. The rules are mine. The discipline is the machine's.

Scaling isn't a heroic move you make once. It's a small, boring decision you make correctly, over and over, on schedule. Set the rules, protect the learning, refresh the tired ads — and let something tireless keep the schedule for you.