Open this doc once per campaign, when Gate 2 is green. This is the single highest-stakes one-time event in the optimization lifecycle. Done correctly: 5-day learning phase, then stable Phase 2 performance. Done badly: 2 to 4 weeks of recovery.
When to do this
- Only when Gate 2 (Transition Readiness) is all green (criteria below)
- Only when there’s been no active budget change or major campaign edit in the last 7 days
- Block 45 to 60 minutes for the switch + same-day verification
Before you start
The campaign must clear Gate 2 before you start the switch steps.
🔴 GATE 2: Transition Readiness
All 4 criteria must be green before switching. Source:,.
| Criterion | Threshold |
|---|---|
| Conversion volume | 15 to 30+ conversions accumulated |
| CPA stability | Weekly CPA within ±25% of monthly average |
| Search terms maturity | Most weekly search terms are familiar (no big surprises) |
| Daily conversion volume | Consistent (not 0-0-0-5-0-0) |
Stay on Maximize Clicks longer if any of these are true:
- Fewer than 10 conversions in last 30 days
- Conversion rate volatile week to week
- New surprise search terms still showing up daily
- Recent budget changes in the last 7 days
30 to 90 days is normal. Switch on data, not calendar. Some campaigns hit Gate 2 at Day 30; others need Day 90.
Steps (run in order)
-
Read or re-read. This is the most expensive routine to fumble in the entire optimization lifecycle. Source it cleanly before touching anything.
-
Pull the cost-per-conversion history. Campaign Overview → Cost per Conversion line. Look at daily / weekly / monthly views; note low and high points; calculate weekly and monthly averages.
-
Get Google’s recommended Target CPA. Settings → Bidding → “Maximize Conversions” → check “Set a target cost per action.” Note the auto-populated number.
-
Choose the bid strategy variant: Maximize Conversions WITH Target CPA (agency default). Set Target CPA at Google’s recommendation + 10% buffer (e.g., Google suggests 97). You can lower it after 2 to 3 weeks of post-switch data.
-
Screenshot existing device bids before pulling the trigger. Smart Bidding removes manual device bid adjustments. Screenshot them for revert reference if the switch needs to be rolled back.,.
-
Verify conversion tracking is still clean. Every conversion action green (Primary); Conversion Goals = Campaign Specific; last 7 days of conversions look reasonable. Smart Bidding is only as good as the conversion data.
-
Execute the switch. Settings → Bidding → save the new strategy.
-
Communicate the switch to the client.
[Inferred client-comms script]Read this verbatim if it helps:
“Quick update: I switched the campaign to a machine-learning bidding strategy today. Google’s algorithm now sets bids to hit a target cost-per-lead of $[X], in the same range we’ve been hitting manually. The algorithm needs about 5 days to settle and 2 to 4 weeks to fully optimize. CPL may fluctuate during that window. That’s expected. I’m watching it daily and we can switch back if it doesn’t perform.”
🔴 GATE 3: Switch Executed Cleanly
Verify all 5 within 24 hours of the switch:
- ✅ Bid strategy reads “Maximize Conversions with Target CPA: $X” in campaign settings
- ✅ Campaign still Enabled and serving impressions (the switch can sometimes pause a campaign briefly)
- ✅ Conversions still tracking (at least one conversion in 24 hours since the switch; if zero, investigate)
- ✅ No new policy disapprovals triggered by the change
- ✅ Client notified
If any criterion fails, revert immediately to Maximize Clicks: Settings → Bidding → switch back. Then troubleshoot before retrying.
The 5-day learning + 4-week observation period
-
First 5 days post-switch: the algorithm is learning. Bids and CPL fluctuate. Do not adjust anything. The temptation to “fix” things during this window is strong. Resist it. Every adjustment resets the learning phase.
-
Days 5 to 30: observation. The algorithm should stabilize. CPL should land within 10% to 20% of your target by Day 14, within 5% to 10% by Day 28.
-
Day 30: if CPL is still 30%+ above target after 4 weeks:
-
Lower Target CPA by 10% and let it run another 2 weeks, OR
-
Revert to Maximize Clicks and try Maximize Conversions (no target) instead, OR
-
Revert to Maximize Clicks permanently. Not every campaign performs better on Smart Bidding.
Watch out for
- Don’t switch mid-budget-change. If you raised or lowered budget in the last 7 days, wait until the campaign re-stabilizes before switching.
- Don’t switch with bad conversion tracking. Smart Bidding is only as good as the data. If anything is off, fix it first (re-run step 6 of the verification).
- Don’t tweak during the 5-day learning phase. Resist every urge to bid-adjust. Each tweak resets the learning.
- Don’t take Google’s recommended target as gospel. It’s usually accurate but not always. Cross-check against your own 90-day average. Use the 10% buffer for safety.
Done when
- ✅ Gate 2 was green before starting
- ✅ Gate 3 passed within 24 hours of the switch
- ✅ Bid strategy now Maximize Conversions with Target CPA
- ✅ Client communicated to
- ✅ Calendar reminders set for Day 14 and Day 28 post-switch reviews
- ✅ Switch logged in
clients/<client>/optimization-log.md
Where to go from here
- Tomorrow: optimization-daily.md (the daily rhythm continues unchanged)
- End of this week: optimization-weekly.md (weekly continues; pay extra attention during the 5-day learning phase)
- Day 30 post-switch: optimization-monthly.md Variant B (the Smart Bidding variant). Variant A is now obsolete for this campaign.