Pulsar
field guide · scaling
Present: ← → Space · Esc back to Doc

Pulsar · Field guide · No. 24 · Why big budget jumps reset learning, vertical vs horizontal scaling, and keeping the creative pipeline ahead of fatigue.

Scale the Winner Without Breaking It

The fastest way to kill a winning ad is to scale it wrong. Grow budget in a way that doesn’t reset learning — and know when to go wide instead of up.

01

Why scaling kills winners

A big budget jump resets the learning phase. The ad set re-enters exploration, pricing spikes, and the winner you were so proud of suddenly looks broken — because you knocked it back to day one (No. 12).

Doubling the budget overnight doesn’t double the winner — it restarts it.
02

Vertical: raise the budget gently

Grow the existing winner in small steps — roughly ~20% every few days — so the ad set keeps its learning instead of resetting. Patience compounds; greed resets.

03

Horizontal: duplicate into new audiences

The other direction: keep the budget-per-set modest and duplicate the winner into fresh audiences. You grow total reach without over-pressuring one pool — which also holds off saturation.

MoveWhen
Vertical (raise budget)the audience is big and still fresh
Horizontal (new audiences)frequency is climbing — the current pool is getting tired
04

Let CBO route the money

Campaign Budget Optimization sets the budget at the campaign level and lets Meta flow it to whatever’s converting — fewer manual resets, less of you fighting the algorithm. Consolidate into fewer, better-fed sets (No. 13).

05

Watch frequency as you scale

Scaling spends your audience faster. Rising frequency means you’re re-serving the same people — the signal to go horizontal (widen) rather than push more budget into a saturating pool.

i
Frequency past ~2.5 with a softening CTR while you scale = saturation, not a bad creative. Widen the audience.
06

Refresh creative before it fatigues

More spend burns creative faster. The winner that carried you at £50/day will fatigue at £500/day. Keep the creative pipeline full — new hooks queued (No. 22) and tested (No. 21) before the current one tires.

07

The scaling routine

Confirm a real winner (enough conversions, stable CPA — not a lucky day).
+20% vertically, or duplicate into a new audience — pick by frequency.
Hold edits. Let it settle before the next move.
Watch frequency + CPA; queue the next creative before this one fatigues.
1 / 8