The Performance Max debate is over. For most e-commerce brands and agencies it isn't a test anymore — it's the default campaign type, and it often does work: it finds liquidity humans miss and scales spend faster than manual campaigns ever could.

It charges for that in transparency. The pitch — "give us your assets, budget, and CPA target, we'll do the rest" — is exactly what should worry an ad ops team. When PMax is winning you don't know why. When it tanks you don't know which lever moved. You're a passenger in your own car.

The mistake is trying to manage PMax inside PMax. Google's interface is built to hide the mess, not surface it. Running PMax safely at scale takes an external layer of governance — watchdog workflows that audit the box from the outside and shout when the algorithm wanders.

What "set and forget" actually costs you

Leave PMax unsupervised and three failure modes show up reliably. It cannibalizes your brand: PMax loves cheap conversions, so it bids into your own brand terms and inflates ROAS with customers who'd have bought anyway. It bleeds into junk inventory: Google has enormous mobile-app and display supply that produces clicks and zero value, and PMax keeps spending there until something stops it. And it hides creative rot behind "Asset Strength," a black box that can score assets "Excellent" while they quietly lose money. None of that means turn PMax off. It means govern it.

Three watchdog workflows that keep PMax honest

Since Google removed the granular controls, you govern PMax by monitoring the data it does release. These are the three we implement first.

Catch brand cannibalization

The signal: PMax ROAS reads 8.0 while blended revenue is flat — PMax is taking credit from brand search. A daily workflow compares brand-search impression share against PMax spend; when brand volume drops as PMax spend climbs, that's the flag, and the strategist gets pinged to check Insights and consider negative brand lists or a rep-side brand exclusion.

Fence in URL expansion

Final URL Expansion is great for reach and terrible for control — it sends traffic to your Terms of Service page, a blog post from 2018, an out-of-stock product. A weekly pull of the PMax landing-page report, filtered to anything that isn't a product or key category page, surfaces those URLs for immediate exclusion.

Kill the mobile-app placement bleed

PMax can dump a fifth of your budget into mobile game apps full of accidental clicks, wrecking CPA. A weekly scan of the placement-detail report, filtered for high spend with zero conversions in mobile-app categories, generates a negative placement list and queues it for approval.

Why this can't just be a manual habit

You could technically check these reports by hand. In practice you won't — Google buries them five menus deep, and doing that across ten accounts is hours you don't have. By the time you find the wasted spend, it's already spent. That's the gap pi-automate fills: we implement these watchdog workflows for you, hit the API for the data that's hard to find in the UI, run the checks, and hand you the result. We don't try to outsmart the bidding algorithm — we build the window that lets you see what it's doing, so you can sleep at night.

Take back control

PMax is a powerful engine. Every engine still needs a governor. If you want one on your accounts, we can stand up a PMax watchdog workflow for you in under 48 hours.

Performance Max is a Black Box. Here Is How to Build a Window.