Performance Max (PMax) for Ecommerce: The Complete 2026 Guide — Asset Groups, Audience Signals, Brand Exclusions & Mistakes

Let's start with a fact a lot of people still haven't accepted in 2026: Performance Max is no longer the set-and-forget black box it was. Over the last two years Google has shipped a pile of controls and reports — negative keywords, brand exclusions, channel-level reporting, asset-level actual metrics, a Final URL report. The box has been pried open. But once you look inside, you hit the other truth: a lot of what you can now see, you still can't change. This guide is the 2026-current version of how ecommerce should actually run PMax.

Google changes this product constantly. Feature names, entry points, and limits move around all the time, so treat every number and button location here as directional and verify in your live Google Ads account and official docs. The underlying logic, though, hasn't really changed: hand bidding, placements, audiences, and budget allocation to the algorithm, and focus on what you feed it — assets and data.

This is the scaling companion to the complete Google Ads for ecommerce guide; the sibling piece is PMax vs Standard Shopping: how to choose.

What PMax actually is

Performance Max (PMax) is Google's AI-driven campaign type: you give it a goal, conversion data, assets, and a budget, and it automatically decides which channel, which person, what bid, and which creative to serve. From one campaign it reaches all of Google's inventory — Search, Shopping, Display, YouTube, Discover, Gmail, Maps — with your product feed as the core engine.

Think of PMax as an automated conversion machine that runs across all of Google's inventory: you give it a goal (usually purchases / conversion value), conversion data, assets, and a budget, and it decides which channel, which person, what bid, and which creative to serve.

It can reach nearly the entire Google ecosystem from one campaign: Search, the Shopping tab, Display, YouTube, Discover, Gmail, Maps, and more. That's a fundamentally different model from Standard Shopping, which only appears in a few high-intent surfaces (Search, Shopping tab, Images).

For ecommerce, the real engine under PMax is your product feed. A bad feed sinks even a smart algorithm — so before you touch PMax, get your Merchant Center set up properly and your feed optimization solid. That's the foundation.

Asset groups: the basic unit of PMax

PMax has no traditional ad groups or product groups. Instead you have asset groups, each typically holding three things:

• Assets: images, logos, headlines, descriptions, videos (if you don't upload a video, Google auto-assembles one from your images — quality varies, so make your own when you can).
• Listing groups: which products / collections this asset group promotes.
• Audience signals: the algorithm's starting hint (more below).

Practical rule: split asset groups by theme / category / margin tier — don't dump every SKU into one group. A "skincare" group and a "makeup" group, for example, so the creative and copy actually match the products. Within each group, feed plenty of varied assets (multiple image sizes, headlines with different angles) to give the algorithm room to test.

Note: each asset group has a cap on listing groups (reportedly in the low thousands), but piling on too many tends to hurt performance. Verify the current cap in Google Ads / official docs.

Listing groups: controlling which products show

Listing groups are one of the few places in PMax where you get product-level control. You can carve products by feed attributes (product category, brand, item ID, custom labels, etc.) and route them into different asset groups.

The most common ecommerce play is to tag products with custom labels (by margin, by sell-through, by new vs evergreen), then use listing groups to isolate your high-margin / fast-moving core products and prioritize them. How to set custom labels: see the feed optimization guide.

Audience signals: a hint, not targeting

This is the biggest beginner misconception. Audience signals are not audience targeting. Your custom intent, customer lists, remarketing lists, Google audiences — these are just starting hints that tell the algorithm "begin roughly with this kind of person." Once it warms up, it can and will serve well outside your signal.

So: make your signals accurate (especially your own first-party customer lists and high-value remarketing lists), but don't expect them to lock the audience. Truly locking an audience and controlling cost is the job of Standard Shopping + Search, not a PMax strength. And first-party data only flows in if your conversion tracking / GA4 pairing is set up correctly — that matters more than anything.

Search themes: giving the algorithm direction

Search themes let you proactively tell PMax which query directions this group should cover — especially useful for long-tail your feed doesn't catch, or new products with no data yet.

• They're directional hints, not exact-match keywords. Don't treat them like Search-campaign keywords.
• There's a cap on themes per asset group (Google's side has commonly been cited around 25 per asset group; the "raised to 50" headline mostly came from Microsoft Advertising, so don't conflate the two — verify Google's current cap in Google Ads / official docs).

The controls you actually have (2026 — there are more now)

People used to call PMax a black box; in 2026 that's clearly loosened. The controls you'll likely have access to now include (confirm in your account):

• Brand exclusions: based on brand entities, not typed keywords — you're excluding brands Google recognizes in its database. The most common use is to exclude your own brand so PMax doesn't claim credit for branded traffic that would convert cheaply anyway (directly tied to the "should PMax eat brand terms" mistake below).
• Negative keywords: now self-serve at the campaign level in the UI, with a high cap (figures in the thousands-plus range have been published — verify the current cap in Google Ads / official docs). Key catch: PMax negative keywords only apply to Search / Shopping inventory — they don't touch Display, YouTube, Gmail, or Discover.
• Account-level placement exclusions: build one centralized exclusion list (sites, apps, YouTube channels / videos) that auto-applies across PMax and other campaign types, for brand safety and to stop wasting spend.
• Device / age / gender exclusions: campaign-level exclusion of certain devices, age ranges, or genders has been rolling out.
• URL-contains rules: feed-based PMax can target / exclude page categories by URL substring (e.g., only pages whose URL contains a given category).

Google moves these controls almost every quarter, so whether they exist, where they live, and their limits — go by your live account and verify in Google Ads / official docs.

Reporting: you can see more; you can't necessarily change more

PMax reporting genuinely improved in 2026, but separate "can see" from "can change":

• Channel-level reporting: the Insights page shows how much budget went to Search, Shopping, Display, YouTube, Discover, Gmail, Maps, and how each performed.
• Asset-level actual metrics: creatives moved from vague "Best / Good / Low" labels toward actual numbers — no more guessing what "good" means.
• Final URL report: breaks spend, conversions, and ROAS by landing page (sliceable by campaign / asset group), great for landing-page optimization.
• Search terms report: it exists, but granularity and depth still trail Standard Search — don't expect the same transparency.

The key black-box residue: you can now see that "Display ate half the budget," but you may not be able to directly cut Display's spend — plenty is still see-it-but-can't-change-it. That's the most frustrating part of PMax in 2026, so manage expectations.

To turn "money you can see" into "money you should spend," first nail your break-even ROAS / breakeven point — don't go by feel. Use our tools to back out gross margin, returns, and ad cost so you know where your red line is; only then does the channel report mean anything.

The mistakes ecommerce makes most

1. Letting PMax eat brand terms by default — without a brand exclusion, PMax tends to claim credit for branded searches that would convert cheaply anyway, inflating ROAS and faking incrementality. Default to adding a brand exclusion and let Search / Standard Shopping handle brand traffic.
2. One giant catch-all asset group — every SKU and category in one group means creative doesn't match products and the algorithm can't learn. Split by category / theme.
3. Running PMax on a weak feed — the feed is the engine; if titles, images, and attributes are off, PMax just burns money. Fix the feed first.
4. Treating ROAS as the only KPI without excluding brand — platform-reported ROAS includes brand lift and attribution fluff. Review incrementality and new-customer share, not just the platform number.
5. Launching before conversion tracking is correct — PMax runs on signal; bad signal means blind spend. Calibrate conversion tracking + GA4 first.
6. Set and forget — PMax is not maintenance-free. Refresh assets, read asset-level actuals, tune creative by channel report, and review exclusions regularly.

Should you run PMax? (the short version)

• New store / very little conversion data: be cautious. PMax needs data; with weak signal it can't learn and tends to overspend. Run Standard Shopping + Search first to generate data and find winners.
• Established winners + creative capacity + enough conversion volume to feed the algorithm: good fit for scaling with PMax.
• Tight budget, need tight cost control: PMax isn't your first choice; Standard Shopping's control suits you better.

For the full trade-off and how to pair it with Standard Shopping (the hybrid setup), see the sibling piece PMax vs Standard Shopping 2026.

Frequently asked questions

Is PMax always better than Standard Shopping? No. The 2026 consensus is "it depends," and many mature sellers run both in a hybrid: PMax for scaling and discovery, Standard Shopping for brand terms, new products, and core-margin items that need tight control. See the sibling article.

Can I make PMax run only Search and not Display? Not with Search-campaign-level granularity. You can see per-channel spend, but directly turning off a channel is limited. Google is testing some channel / partner-network selection — go by your live account and verify in Google Ads / official docs.

Will search themes waste money? They're directional hints, not keywords. Pair them with negative keywords + brand exclusions to contain some of the waste on the Search / Shopping side.

How often should I check it? At least weekly: asset-level actuals, channel report, search terms, negatives and exclusions. Keep assets on a rolling refresh.

Bottom line

2026 PMax: the black box has a crack in it, but "visible" doesn't equal "controllable." Winning with it for ecommerce isn't about dashboard buttons — it's about the three things you do control: a clean feed, on-point creative, and accurate conversion data + first-party signals, plus excluding brand by default, splitting asset groups by category, and a weekly review. Do those well and PMax is a scaling weapon, not a money pit. For every specific limit and feature location, defer to your Google Ads account and official help docs.