The short, defensible answer: increase Google Ads budgets by no more than 20% at a time, and wait 7-14 days between increases. Bigger jumps push campaigns back into the learning phase, reset Smart Bidding's confidence, and almost always cost you efficiency for 2-3 weeks. Below, Ben breaks down when 20% is too cautious, when it's too aggressive, and how to read the signals that tell you which.
The short answer
For most UK accounts running Smart Bidding (tCPA, tROAS, Maximise Conversions, Maximise Conversion Value): +20% per change, every 7-14 days. For manual CPC search campaigns with stable Quality Scores: you can move faster, often +30-50% in one step. For Performance Max: stick to 15-20% and give it the full 14 days because the algorithm rebalances across asset groups, audiences and channels.
The 20% rule (and why it exists)
Smart Bidding learns by spending. When you change a daily budget by more than ~20%, the system treats it as a meaningful change to the auction conditions and recalibrates. In practice that means a 7-14 day "learning" period where CPA and ROAS swing wider than usual. Stay under 20% and the recalibration is so small you barely see it.
This isn't a Google rule written down anywhere - it's a pattern Ben and most senior practitioners have observed across hundreds of accounts. The threshold isn't magic at 20.0%; it's just a safe ceiling that almost never triggers a destabilising re-learn.
How often to increase budget
| Account state | Max increase | Wait between |
|---|---|---|
| New campaign, <30 conversions | +10-15% | 14 days |
| Stable, hitting CPA/ROAS target | +20% | 7-10 days |
| Mature, 100+ conv/mo, beating target | +20-30% | 7 days |
| Manual CPC search, stable QS | +30-50% | 3-7 days |
| Performance Max, mature | +15-20% | 14 days |
Signals you're ready to scale
Don't just increase budget on a hunch. Scale when you see at least two of these:
- Lost IS (budget) above 15% - the campaign is being capped before the day ends.
- CPA or ROAS beating target for 14+ consecutive days.
- Search impression share at or above 70% with budget headroom (lost IS rank low).
- Stable conversion volume week-on-week (not a spike from a one-off promotion).
- Quality Scores holding at 6+ across top spending keywords.
Warning signs you're scaling too fast
- CPA jumped 30%+ in the week after an increase - and stayed there.
- Conversion volume didn't move despite a meaningful budget lift.
- "Limited by budget" replaced by "Learning" status that won't exit.
- CPC creeping up while conversion rate falls - you're buying lower-quality auctions.
If you see any of these, hold the budget flat for 14 days and let the system restabilise before the next move.
Increase steps by campaign type
Search (Smart Bidding): +20% every 7-10 days. Watch impression share and CPA together.
Search (Manual CPC): +30-50% in a single step is fine if Quality Scores are stable - manual bids don't relearn.
Performance Max: +15-20% every 14 days. PMax rebalances spend across Search, Shopping, Display, YouTube and Discover, so it needs more time.
Shopping (standard): +20-25%. Less sensitive than PMax because there's no creative rotation.
Display and Video: +20%. CPMs are noisy week-to-week, so give it 14 days before judging.
Demand Gen: +15-20%. Treat it like PMax - the algorithm needs time to rebalance placements.
Scaling under Smart Bidding
Two budget mistakes that wreck Smart Bidding accounts:
- Doubling daily budget overnight. Even on a profitable campaign this triggers a learning event - expect 7-14 days of swinging CPA before it settles. Ben has rescued multiple accounts where the previous manager doubled budget for a "big push" and ROAS halved for three weeks.
- Increasing budget while also tightening tCPA/tROAS targets. Two opposing instructions at once. Change one variable at a time, wait, then change the next.
When a big jump is actually fine
The 20% rule has exceptions. You can safely make larger jumps when:
- Brand campaigns - search demand for your brand name doesn't care about Smart Bidding learning.
- Manual CPC with stable Quality Scores - no algorithm to retrain.
- Seasonal events you've prepped for (Black Friday, Boxing Day) - if you've raised budgets gradually in the run-up, a final 50% push on the day is fine.
- You're moving budget between campaigns rather than increasing total account spend - the auction footprint barely changes.
Worked example: £2k to £10k/mo
A UK home services client wants to scale from £2,000/mo (£66/day) to £10,000/mo (£330/day). Five-fold increase. Done in one move it would tank performance for a month. Done properly:
- Week 1: £66 → £79/day (+20%). Hold 10 days.
- Week 3: £79 → £95/day (+20%). Hold 10 days.
- Week 5: £95 → £114/day (+20%). Hold 10 days.
- Continue +20% every 10 days until £330/day.
Total time to 5x: roughly 9-10 weeks. Painful if you wanted overnight scale, but you keep CPA in target the entire way - which on this account meant an extra ~£14k of profitable revenue versus the "double it now" approach that produced 3 weeks of unprofitable spend.
