Blog

When and How Much to Scale Ad Budget — The Scaling Rules

ImageFactory Engineering · Published 2026-06-18

You've probably had this happen: a winning ad is doing great, you push the budget up hard, and the next day efficiency falls apart. It isn't bad luck. Meta and Google's machine learning treats a big budget change as "the conditions changed" and restarts its learning from scratch. So budget scaling has one rule worth memorizing — never raise more than 20% at once. And a second one people miss — raising the budget burns through creative faster. Scaling is ultimately a "budget + creative supply" game. Let's start with the timing of increases and cuts.

How much can I raise the budget at once?

20% is the ceiling. Raise a daily budget by more than 20% in a single edit and the algorithm re-enters the learning phase, spiking CPA and dropping ROAS in the meantime (the 20% rule and learning reset). Big audience, creative or optimization-event changes reset it the same way. So the standard move is +20% every 2–3 days.

When should I increase it? (timing)

Check three things before you raise.

1. Has it left learning? — has the ad set hit roughly 50 conversions a week to exit the learning phase.
2. Is performance stable? — if day-3 CPA is within 20% of the start and day-7 ROAS within 15%, that's a green light.
3. Is frequency rising too fast? — if frequency is spiking, increasing budget only accelerates fatigue.

The more urgent it feels, the slower you should go. Waiting 3–7 days to confirm the signals and raising 20% at a time is, in the end, the fastest path.

Vertical vs horizontal scaling

There are two directions to grow spend.

• Vertical (raise the budget on a winning ad set): fast, but it saturates one audience quickly.
• Horizontal (duplicate the winner into new audiences, lookalikes and placements): more headroom.

The rule of thumb is "push vertically until frequency or CPA pushes back, then go horizontal". When moving to a CBO, don't 5× it — start at the combined spend of your winners and scale ~20% at a time, or you reset learning across the whole campaign.

If ROAS drops, should I cut the budget?

A hasty cut costs you. Check frequency before cutting. If 7-day frequency is above ~2.5 and climbing, it's creative fatigue, not a bad audience — so refresh the creative or widen the audience instead of dropping spend. The 2.0–2.5 band is a warning zone; swap creative before you reach it.

Why does raising the budget burn creative faster?

This is scaling's hidden bottleneck. A bigger budget shows your ads to the same people more often, so frequency rises, CTR falls and fatigue sets in. At high spend you need 5–10 fresh creatives a week to keep the scale from breaking within 1–2 weeks. In other words, to earn the right to raise budget, you have to be able to supply creative at that pace.

Scaling checklist

1. +20% max, every 2–3 days.
2. Before raising: learning done + CPA/ROAS stable + frequency checked.
3. Push vertically; when it resists, go horizontal.
4. On a ROAS dip, check frequency first — if it's fatigue, swap creative, don't cut.
5. The higher the spend, the more you need 5–10 fresh creatives a week ready.

How ImageFactory helps here

The thing that stalls scaling is usually not budget but creative supply — shooting and designing 5–10 new pieces a week by hand isn't realistic. ImageFactory turns one product cutout into multiple style variations and per-platform size adaptations automatically, feeding the algorithm fresh creative without a designer bottleneck. Having the next creative ready before frequency climbs is the most practical way to protect a scale.