Consolidate or segment? A Performance Max decision framework

Few Google Ads campaign types divide PPC practitioners quite like Performance Max.

Some advertisers consolidate everything into Performance Max and let the machines run the show. Others treat it as a black box that eats their budget and avoid it altogether.

After running Performance Max across dozens of accounts, I can tell you that neither extreme is the right approach. Whether you lean into it or shut it off should depend on the signals in your account, not your ideology.

The same account that should consolidate today might need to diversify six months from now. Brand cannibalization, feed quality, channel performance, conversion volume, and product margins should guide those decisions.

The two extremes of Performance Max

Before we get into the signals, it helps to see the two traps that most of the accounts I audit are sitting in.

Excess consolidation

This usually comes from taking Google’s recommendations at face value. One campaign ends up owning the whole account (and budget), and because Performance Max is algorithmic, you can’t control where the spend goes or what gets promoted over something else. 

The classic version is branded Search getting cannibalized by Performance Max. Your blended ROAS looks great, but a big chunk of that demand was already owned. Google can’t tell the difference between a brand conversion and a net-new one, so it takes credit for a 10x ROAS while your actual prospecting efficiency suffers.

Excess isolation

This usually comes from carrying forward the old ways we used to run Shopping and Search. You build 15 campaigns split by every product line, audience, and one-off idea. The problem is that Performance Max needs data volume to succeed. Fragmentation means no campaign gets enough conversion data to exit learning, so everything sits half-optimized while the budget gets spread too thin.

Dig deeper: The parts of Performance Max you can actually control

5 signals to guide your Performance Max campaign structure

Excessive consolidation and isolation have the same root cause: running Performance Max with the mindset of an older campaign type rather than what it actually is. But the five signals below are how you figure out where your account sits on the spectrum between these extremes.

Signal 1: Brand traffic cannibalization

What this signal tells you: Whether your reported ROAS is real or inflated.

The first thing I check in any Performance Max campaign is the brand vs. nonbrand split. If it’s eating up branded searches, you’ll see it in the search term insights and in the impression share of your brand terms.

The inability to set separate bids or targets for branded and generic terms means that when Performance Max grabs a brand conversion, the system sees a cheap and easy sale … and then decides this is what good looks like for your account. It then goes after more conversions at that cost level, which inflates your CPCs on branded terms you could win for a fraction of the price.

The double penalty is that prospecting efficiency gets masked, and you pay a premium to convert people who were already likely to buy from you.

Excluding brand universally isn’t the answer because Performance Max is a full-funnel campaign by design. Doing this denies the system actual signal about how someone travels from awareness to purchase. The decision has to be conditional.

• For most brands, cost efficiency matters. You’ll want to run a dedicated branded Search campaign on Target Impression Share with exact or phrase match to show up without overpaying.
• For a very large advertiser who cares more about full-funnel signals than squeezing incremental CPC gains, leaving it in can be the right call.

Signal 2: Feed quality and readiness

What this signal tells you: Whether Performance Max even has a fair shot.

For ecommerce accounts, Performance Max is only as good as the feed underneath it. Miscategorized products, nondescriptive titles, and missing attributes mean that Google is optimizing based on bad data. No amount of budget fixes that.

If your feed is broken, the first move has to be to fix it. Don’t throw a broken feed into Performance Max and expect Google to do anything other than find the path of least resistance.

Dig deeper: How campaign structure shapes Google Ads performance

Signal 3: Channel-wise ROAS

What this signal tells you: What actually works and what’s being carried.

Performance Max reports one blended ROAS across Shopping, Search, Display, and Video. If one channel is crushing it and another is lagging behind, the surface value can mask the economic story.

• Turn off URL expansion so you stop funneling budget into low-intent Display placements.
• Look at the channel performance breakdown to know exactly how much budget is going to feed-based ads versus Video and Display.
• If you’re getting traffic from Search and don’t want any, remove search themes. This will force it to focus on Shopping inventory, listed in the channel performance breakdown as “Google Search – Ads from product data feed.”

Then, it’s all about structure.

If specific products or segments aren’t getting traction inside Performance Max, I tend to exclude them and test them in Standard Shopping. You aren’t trying to pick one structure and force every account into it. Test which setup wins for a given account and put the budget where you can see that it makes money.

Signal 4: Account size and data volume

What this signal tells you: Whether you have the data to support more campaigns.

I tend to argue that campaign count should follow data volume, not your creative whims or your product catalog.

Small accounts can’t usually fund five campaigns, whether due to budget or demand volume (or both). If there aren’t enough conversions to go around, the answer is to go simple. Get everything into fewer, more important campaigns so the system has enough budget and signals to optimize.

For large accounts with bigger product ranges, one campaign won’t give you category-level budget control or protect your high-value categories from being averaged in with everything else. That’s when splitting into a handful of campaigns by product category makes sense.

One important caveat: I see people build multiple asset groups targeting different audiences with the same products, which goes against how Performance Max works. 

You can’t target a fixed audience, so slicing it that way just confuses the system and gives you nothing meaningful in reporting. Build around categories, matching each campaign to a specific business goal with its own purpose. Then let the audience signals do their job one by one.

Signal 5: Category and margin structure

What this signal tells you: Whether segmentation helps or hurts.

Conversion data and demand are vital, but they’re not everything. Even when there’s plenty of both, segmentation prevents high-value SKUs from being averaged in with products that have very different economics (margin, price point, buying cycle).

A flagship line that’s regularly sold out (60% margin) and a new accessories range being tested (15%) don’t belong under a single campaign with a blended target. Google will trade the former for the latter and report it as a win.

Dig deeper: Google tests channel prioritization controls for Performance Max

Keep it flexible

These five signals need to be viewed as ranges, not absolutes. No single signal outweighs all the others, and no decision is permanent. You keep monitoring the account’s health and depth against the direction of the business.

The same account that needs to get off Performance Max today may warrant the opposite call six months later once the feed is cleaned up, conversion volume doubles, or a new category launches with new margins.

And when the decision framework guides you to wall things off, don’t cut everything at once.

A huge shock to the algorithm can reset learning, disrupt attribution midcycle, or take out a campaign that was feeding your funnel. Cut in phases, allowing the account time to restabilize between moves. And avoid doing anything major heading into a sales cycle or holiday window.