How to diagnose Performance Max campaigns with an AI agent
Where PMax spend actually goes, whether asset groups are healthy, and which campaigns spend without the ad strength to justify it.
IntentMake an opaque campaign type inspectable: placements, asset health and spend, in one ranked view.
- 5
- Steps
- 5
- Tools
- 4
- Rules
- 3
- Failure modes
When to use it
Use it when
- PMax holds meaningful spend and nobody can say where it goes.
- Before feeding PMax more budget.
- Shopping accounts where PMax and standard Shopping coexist.
Do not use it when
- PMax campaigns younger than 6 weeks: the learning phase makes every diagnostic read as unhealthy.
- Accounts where PMax spend is trivial; the pass costs more attention than it frees.
What it needs first
Required
Preconditions
- At least one PMax campaign with 30+ days of spend.
- For product questions, the Shopping feed reachable via the account.
Inputs
window- Analysis window.Default: last 30 days.
Procedure
In order. Every tool named is one the gateway ships, and links to its reference.
Read asset group diagnostics: ad strength, coverage, action items.
Read the assets themselves, to name what is missing or weak.
Read placement distribution: how much serves on Search-like, Shopping, Display, YouTube.
For Shopping-heavy accounts, read product-level performance and the products behind it.
Rank campaigns by spend over asset health, and list the top concrete fixes.
Decision rules
The observable condition, what it lets you conclude, and what takes the conclusion back.
When
A campaign spends above the account median with ad strength below Good
Conclude
Rank it first: it buys reach with weak ammunition, and the fix is asset work, not bidding.
When
Coverage action items name missing asset types (video, long headlines)
Conclude
List them verbatim as the work order.
When
A high share of conversions comes from brand-like placements
Conclude
Flag possible brand cannibalisation; PMax may be buying conversions search would have caught.
Unless
Brand is explicitly excluded elsewhere and this is the intended catcher.
When
Products with spend and zero conversions concentrate in one feed segment
Conclude
The problem is the feed segment, not the campaign; name the segment.
Evidence every conclusion must carry
- Every ranked campaign shows spend, ad strength, and its named action items.
- Placement claims cite the placement report, not inference.
Where the agent stops
- Asset replacement and feed fixes are human work; the skill delivers the ordered list.
- Budget or status changes proposed here go through Safe Writes.
How it goes wrong quietly
The cases where the analysis is wrong and still looks right. Read them before trusting a number.
- Learning-phase campaigns read as broken; check campaign age first.
- Placement reporting for PMax is partial by design; treat distributions as indicative, and say so.
- Ad strength is Google's own heuristic; treat it as a ranking signal, not a verdict.
What the answer contains
- 01Campaigns ranked
- Spend over asset health, worst first.
- 02Work order
- Missing or weak assets, verbatim from diagnostics.
- 03Spend map
- Where PMax money actually serves, with the caveat stated.