To scale Facebook ads without resetting the learning phase, raise budgets gradually, around 20% at a time, a few days apart, and expand horizontally with new audiences and creatives instead of making big edits to a winning ad set. Only scale ad sets that have already exited learning.
What the learning phase is (and why it matters)
When you launch or significantly edit an ad set, Meta’s delivery system needs data to learn who to show it to. Meta says an ad set generally exits learning after about 50 optimisation events within 7 days of its last significant edit (Meta Business Help Center).
During learning, performance is less stable and costs are often higher. Significant edits, big budget changes, new creative, new targeting, a new optimisation event or bid strategy, can send it back into learning.
Two ways to scale
| Vertical scaling | Horizontal scaling | |
|---|---|---|
| How | Increase the winning ad set’s budget | Duplicate winners to new audiences, add creatives |
| Pros | Keeps accumulated learning | Finds new demand, lower risk to the original |
| Cons | Large jumps can destabilise delivery | New ad sets start learning from scratch |
| Best for | Steady growth on proven ad sets | Breaking through a plateau |
The 20% rule for vertical scaling
- Pick ad sets that have exited learning and hit your target CPA or ROAS for several days.
- Raise the budget by about 20%.
- Wait 2 to 3 days and check CPA, frequency and delivery.
- If results hold, repeat. If CPA climbs, hold or step back.
The 20% figure isn’t a Meta law, it’s a widely used, conservative practice that keeps changes small enough for delivery to stay stable.
Horizontal scaling ideas
- New audiences: broad targeting, lookalikes of purchasers, new interest seeds.
- New creatives: fresh angles on your winning offer (see creative fatigue).
- New placements or formats: 9:16 versions for Reels and Stories.
- New markets: add cities or countries one at a time.
A simple 4-week scaling plan
| Week | Action |
|---|---|
| 1 | Confirm winners have exited learning; +20% budget on the best two |
| 2 | +20% again if CPA holds; add 2 new creative variants |
| 3 | Duplicate the top ad set into a broad or lookalike audience |
| 4 | Review: keep scaling winners, cut ad sets above target |
Mistakes that reset learning
- Doubling budgets overnight.
- Changing budget, creative and targeting at the same time.
- Pausing and re-launching to “reset” performance.
- Scaling ad sets that are still in learning.
How Spotnxt scales safely for you
Spotnxt’s Automated Rules and Autopilot have the guardrails built in: budgets move at most 20% at a time, with a 48-hour cooldown and learning-phase protection. The Recommendations Dashboard flags room to scale with a ₹ impact estimate, every change is logged, measured a week later and undoable.
See Automated Rules & Autopilot → · Read: Best Facebook ads automation tools →