The module closes with the procedure for looking at an existing PMax campaign —yours or one in an inherited account— and deciding with data whether to scale it, rein it in or switch it off. Two parts: the configuration audit (what should be there and isn't) and the incrementality test (what it really contributes).
The 20-point audit
Measurement
- The campaign's conversion goal = primary actions with value only.
- Real values (excluding VAT, net of returns); offline conversions if it is lead generation.
- Click-through vs view-through conversions kept separate in the reports.
Exclusions (lesson 4)
- Brand exclusions applied.
- Account negatives (universal) and campaign negatives (roots).
- Account placement list applied and alive; apps excluded if they don't apply; content exclusions.
- URL exclusions or expansion turned off where appropriate.
- Excluded audiences (recent customers if the goal isn't retention).
Structure (lesson 2)
- Asset groups by audience/family; 4-6 at most.
- Complete creatives (every aspect ratio, your own video).
- Signals = your data + custom segments from your core.
- Automatically created assets reviewed.
- Product listing disjoint from standard Shopping (lesson 5).
Bids and budget
- A strategy with a target (CPA/ROAS) set from the real figure; not "Maximize" with no brake and a generous budget.
- Budget in line with the IS and the marginal return (module 9 of the Intermediate level).
Reporting (lesson 3)
- Channel performance reviewed: no channel with high cost and zero click-through conversions.
- Search terms: brand out, negatives up to date.
- Placements: no recognisable junk.
- Assets: no "Low" ones with volume left abandoned.
- Change history: nothing touched in the last 2-3 weeks before judging.
The incrementality test
The question: are PMax's conversions new, or did they move over from Search, Shopping and brand? Three methods:
1. Official experiment
Google offers experiments for PMax (against standard Shopping, or "PMax on vs PMax off" with a traffic split). Several weeks, statistical significance. It is the clean method.
2. Controlled pause
Pause PMax for 2-3 weeks (out of season) and measure the account's total conversions against the previous weeks (and the same period the previous year). If the total falls by what PMax was contributing, it is incremental; if the total doesn't move and Search/Shopping go up, it was cannibalisation. The cost: the risk of losing volume during the pause.
3. Geo experiment
PMax on in some regions and off in others, 4-6 weeks, total sales by region. For large accounts with locatable sales.
A complement to all three: compare with brand exclusions already applied (otherwise the result is contaminated).
The decision
| Result | Decision |
|---|---|
| Incremental with CPA/ROAS on target | Scale in steps (budget, more groups, more families) |
| Incremental but expensive | Rein it in: more exclusions, a tougher target, strip out junk channels via placements, move low margin to standard |
| Barely incremental (it was moving conversions) | Reduce it to what does contribute (e.g. prospecting only, with new signals) or switch it off; give the budget back to Search/Shopping |
| Not enough data | Keep it reined in; repeat the test with more volume |
And in every case: the 20-point audit before the test. Testing the incrementality of a badly configured PMax campaign only proves that a badly configured PMax campaign contributes nothing.
💡 Ninja trick: half the points in the audit can be watched every night: the Agent reviews the conversion goals and exclusions of every PMax campaign, Ninja Shield maintains the account placement list and Shopping Ninja the split of products by margin. The incrementality test is still a human decision — but with the configuration already clean, the result is credible.
What you should remember
- Audit 20 points (measurement, exclusions, structure, bids, reporting) before judging.
- Incrementality with an official experiment, a controlled pause or a geo test; always with brand exclusions in place.
- Decision: scale, rein in, reduce/switch off depending on incrementality and cost.
- A badly configured PMax campaign never proves anything.
That is the end of Module 1 of the Advanced level. Module 2 is Demand Gen: what it really is, creatives, audiences, fatigue and measurement.