Scaling a campaign without resetting the learning phase

Why big jumps break campaigns

Every bidding system needs recent conversion data to price the next impression. When spend triples overnight, the system starts buying audiences it has never tested, and the data it learned from stops matching reality. Performance drops, the team panics, and the budget gets cut — which resets everything again.

Scale in increments

A practical rule is 20% to 30% per day on the campaigns that are already profitable. That pace is slow enough for the algorithm to keep learning and fast enough to matter over a month. If a campaign is far ahead of target, allow one larger jump, then hold for three days before increasing again.

Protect the core

Scaling should not mean changing everything. Keep the winning creative, the winning sources and the winning GEOs untouched, and put the increased budget into adjacent tests: a new source, a new angle, a neighbouring market.

• Keep one control campaign at the original budget for comparison.
• Add new sources gradually, with their own caps.
• Cut losers weekly, not daily.

Watch the right metrics

The signal that scaling is working is not higher spend — it is stable cost per event. If cost per event rises more than 15% after an increase, pause the increase and let the campaign settle before trying again.

Scaling is boring on purpose. The campaigns that reach eight figures are usually the ones that resisted the urge to double overnight.