← Writing

Half your Meta campaigns are starving each other of signal

Fewer campaigns is the least popular thing I say on a first call. Everyone would rather hear it was the creative, or the bid, or iOS.

So we pull one column. Optimisation events, per campaign, per week.

In this US account, half the campaigns had never cleared 50 in a week. Those four produced 70 events between them. One campaign’s worth of signal, spread across four.

Half these Meta campaigns never reached 50 events a week. They took 23 percent of the budget and returned 54 percent day-28 ROAS. The other four returned 80 percent. Scatter plot of optimisation events a week, log scale, against ROAS at day 28, bubble size equals spend, with a marker at 50 events a week. Under 50 events a week: 4 of 8 campaigns, $16,843, 23% of budget, 54% ROAS. 50 events a week or more: 4 of 8 campaigns, $58,012, 77% of budget, 80% ROAS. 8 US campaign-months, Adjust cohort ROAS at day 28, April to June 2026. Under 50 events a week: 4 of 8 campaigns, $16,843, 23% of budget, 54% ROAS. 50 events a week or more: 4 of 8 campaigns, $58,012, 77% of budget, 80% ROAS. The four under-50 campaigns all launched mid-May, so age and signal move together; holding the month constant, the split is 75% against 54%, so the gap survives after controlling for launch age.

Nobody sets out to do that. It happens one launch at a time. A campaign for the new platform. A campaign for the test. A campaign for the team, because the last one got messy and starting fresh felt cleaner.

None of it looks like a mistake while you are doing it. Fewer campaigns, same budget, everything clears the line.