The Performance Max Playbook for 2026: Structure, Signals, and Control

What Performance Max is really optimizing

Performance Max is a single campaign type that serves across Search, Shopping, YouTube, Display, Discover, Gmail, and Maps, using one budget and one bidding goal. It optimizes for the conversion action you tell it to optimize for, at the value you report back.

That last sentence is the entire game. PMax is not smart or dumb. It is obedient. If you feed it revenue, it chases revenue. If you feed it profit, it chases profit. If you feed it every form fill on the site, it will happily buy you a warehouse full of junk leads.

The three levers you actually control

You cannot pick placements, you cannot see most search terms, and you cannot bid per keyword. You still control the three things that decide the outcome:

The conversion signal. Which action, and what value is attached to it.
The inputs. Product feed quality, asset groups, audience signals, and listing groups.
The boundaries. Brand exclusions, negative keyword lists, account-level exclusions, and geographic targeting.

Everything else is commentary.

Account structure that works

We use a simple pattern across ecommerce accounts:

One PMax campaign per margin tier, not per product category. Products with a 60% margin should not share a budget with products at 12%.
A separate brand Search campaign with exact match, plus brand exclusions applied on PMax, so PMax cannot claim credit for demand you already had.
A generic Search campaign for the head terms you want explicit control over.
One asset group per meaningful audience or product story, with distinct creative, not the same five images recycled.

For lead generation, replace margin tiers with lead value tiers and make sure offline conversion imports are running before you launch.

Feed quality beats creative volume

In Shopping-heavy PMax, the product feed is the targeting. Titles, product types, and attributes tell Google what queries to match you to.

Practical wins we see repeatedly:

Rewrite titles to front-load the terms buyers actually type, not internal SKU language.
Fill product_type with a full taxonomy path, not a single word.
Add custom_label_0 for margin tier, custom_label_1 for stock depth, custom_label_2 for seasonality. These become your segmentation levers.
Keep GTIN and brand accurate. Missing identifiers quietly suppress impressions.

A clean feed usually moves performance more than any amount of new video assets.

Feed profit, not revenue

The single highest-impact change we make in most accounts is switching the conversion value from revenue to profit. With ProfitMetrics or a server-side value adjustment, the value sent to Google becomes gross profit after cost of goods, shipping, payment fees, and returns.

The effect is immediate and uncomfortable at first. Reported ROAS drops. Actual profit rises, because the algorithm stops buying high-revenue, low-margin orders to hit a target that never reflected reality.

Set your target once you have two to four weeks of profit-based data, not before.

Exclusions are not optional

Before a PMax campaign gets meaningful spend, apply:

Brand exclusions for your own brand terms, unless brand defense is the campaign’s job.
A negative keyword list at account level for competitor names, job seekers, support queries, and free-intent terms.
Content exclusions for sensitive placements and, for most B2B accounts, an exclusion of mobile app inventory.
Geographic targeting set to “presence” rather than “presence or interest”.

That last setting alone cleans up a surprising amount of waste in accounts serving a single country.

Reporting you can actually act on

The insights tab is not enough. Build a Looker Studio view that pulls:

Asset group level spend, conversions, and profit.
Search category insights with week-over-week change.
New customer acquisition rate versus returning.
Product-level profit from your feed joined to Shopping performance.

If you cannot answer “which asset group made money last week?” in under a minute, your reporting is not finished.

A realistic 30-day plan

Week 1: fix the feed, define margin tiers, and switch conversion values to profit. Week 2: rebuild campaign structure by margin tier and apply all exclusions. Week 3: hold budgets steady and let the learning period complete. Resist edits. Week 4: read profit by asset group, cut the losers, and shift budget to the tiers that clear your target.

Performance Max rewards patience and punishes fiddling. Set it up properly, then leave it alone long enough to learn.