Performance Max for Shopify: What to Feed It, What to Exclude (2026)

Seventy-one percent of Google Ads accounts run Performance Max, and most of them run it badly. Not because PMax is bad, but because the default setup does two things that quietly hurt Shopify stores: it cannibalises your branded search traffic and reports the resulting harvested conversions as new performance, and it spreads budget across Display and YouTube before it has learned anything useful about your Shopping demand.

This guide is the version with opinions. How PMax actually works in 2026, the exclusion layers that most accounts never set, asset group structure that does not dilute budget, and how to read the reporting honestly enough to know whether the campaign is growing your store or just claiming credit for demand you already had.

What changed in PMax by 2026

A lot of the "PMax is a black box" criticism is now out of date. The controls that exist today:

• Full search terms reporting, matching Standard Shopping visibility
• Brand exclusion lists, up to 1,000 entries per account
• Account-level negative keywords
• First-party audience exclusions, so you can exclude existing customers from acquisition campaigns
• Channel-level performance reporting showing budget split across Search, Display, YouTube, Discover, and Maps
• Search themes expanded to 50 per asset group (up from 25)
• Asset-level performance ratings (Best / Good / Low) in the Asset Group Performance report
• Search Partner Network placements visible in the "when and where ads showed" report as of February 2026

What you still cannot do: set arbitrary negative keywords at campaign level (only account-level negatives and brand exclusions), see channel splits at asset-group level without the API, or bid manually.

The most important thing: exclude your brand

If you take one thing from this guide, take this one.

By default, Performance Max will bid on searches for your own brand name. Those shoppers were going to buy anyway. PMax wins the auction, pays a higher CPC than a branded Search campaign would, and reports the conversion as PMax performance. Your reported ROAS looks excellent while your blended acquisition cost goes nowhere.

Incrementality testing by Haus found that excluding brand terms from PMax drove around 24% more incremental revenue on average across their tests, and cut new-customer acquisition cost by roughly 40%.

Practitioners running this on client accounts report the same pattern: reported PMax ROAS drops (7x to 4x is a typical first-fortnight move) while blended ROAS climbs (3.8x to 5.2x once things settle). It looks like a downgrade in the dashboard and is an upgrade in the P&L.

How to set brand exclusions

1. Tools > Shared library > Brand lists. Create a list containing your own brand name and its common misspellings.
2. Apply the list at campaign level on your PMax campaign, or account level if you run several.
3. Optionally add competitor brands you do not want to bid on. This has a smaller effect: slight impression drop, slight CPA improvement, showing up over about 30 days.
4. Build a separate branded Search campaign with controlled bids to capture that traffic at a fraction of the cost.

Brand exclusions apply at campaign or account level only. Asset groups inherit from the parent campaign; you cannot set them per asset group. You should see brand query impressions drop within 24-48 hours.

Note that for retailers with product feeds, Google added a Search-only exclusion option, letting you exclude brand terms from Search text ads while still capturing branded traffic through Shopping ads. Whether that suits you depends on whether your branded Shopping traffic is genuinely incremental.

How to verify PMax was eating your brand

Pull a query-level report from your branded Search campaign for the last 90 days. Compare impression share before and after PMax launched. If impression share dropped 15-30 points without any budget cut on Search, PMax was taking it. After setting exclusions, branded Search impression share typically recovers within two weeks.

The feed is the campaign

PMax spends most of its money on Shopping placements. That means your Merchant Center feed, not your ad copy, is the primary performance lever. The feed decides which queries you are eligible for at all; a thin feed caps performance no matter how good the bidding gets.

Feed work that moves PMax performance:

• Titles structured for search. Brand + product type + key attribute + size/colour. Google matches queries against titles heavily.
• Correct Google product category on every SKU. The auto-guess from Shopify product type is wrong often enough to matter.
• GTINs on branded products, identifier_exists: no on private label.
• Complete attributes: colour, size, material, age group, gender where applicable. Missing attributes reduce eligibility for refined queries.
• Custom labels for segmentation. Five slots per product. Use them for margin tier, best seller status, stock level, and price band, because these become your listing group structure.
• High-quality images, 800x800 minimum, no text overlays.

If your feed is not clean, fix it before you touch campaign structure. Feed work compounds; campaign tweaking on a bad feed does not.

Campaign structure: how many campaigns, how many asset groups

Start with one campaign

The instinct to split into many campaigns dilutes budget and slows learning. Start with one PMax campaign covering your full catalogue, with brand excluded. Split only when you have a clear reason:

• Margin tiers. High-margin and low-margin products with genuinely different ROAS targets belong in separate campaigns, not separate asset groups, because ROAS targets are campaign-level.
• Category with distinct economics. A category with a very different AOV or repeat rate justifies its own campaign.
• New customer acquisition. Using first-party audience exclusions to exclude existing customers creates a genuine acquisition campaign, which is worth separating.

Keep asset groups few

More asset groups means more budget splitting and slower learning. Start with one or two focused asset groups and add more only when data justifies it. Use listing group exclusions inside asset groups to control which products each one serves.

Feed-only for the first month

Start new PMax campaigns with feed-only configuration for the first two to four weeks. This runs Shopping-style ads from the feed without additional creative assets, concentrating spend on Shopping placements where ecommerce conversion rates are strongest.

Two benefits: you get a clean Shopping performance baseline before the algorithm starts allocating budget across Display and YouTube, and you avoid the common failure mode where weak creative assets pull spend into low-intent placements during learning.

Add creative assets once you have a stable Shopping baseline to compare against.

Audience signals and search themes

Neither of these is targeting. Both are hints.

Audience signals tell Google where to start looking. The strongest signals are first-party: your customer list, your high-value customer segment, your cart abandoners. Upload these as Customer Match lists. Interest-based and demographic signals are weaker but still directionally useful.

Search themes operate at asset-group level, up to 50 per group. Treat them like exact-match hints. Vague themes ("running shoes") are useless because the algorithm already knows that from your feed. Specific ones ("women trail running shoes wide fit") direct it toward queries you actually want. Five to fifteen specific themes per asset group is the useful range; pull them from your top-converting search terms in existing Search campaigns.

The three exclusion layers most accounts never set

1. Brand exclusion list. Covered above. Your own brand, always. Competitors, optionally.
2. Account-level negative keywords. Terms that consistently produce zero conversions across your account. Job-seeker queries, "free," "cheap," "how to make," competitor product model numbers you do not stock, and anything else the search terms report shows as pure waste.
3. First-party audience exclusions. Exclude your existing customer list from acquisition campaigns so you are not paying to reacquire people who would have bought anyway. This is the 2026 feature that solves PMax's most persistent problem.

Additionally, use listing group exclusions to keep PMax and Standard Shopping in separate product lanes if you run both. Without this you are bidding against yourself.

Bidding: how to actually run it

PMax uses Smart Bidding exclusively. Two strategies:

• Maximise Conversions / Maximise Conversion Value without a target. Use this at launch, while the campaign builds a conversion model.
• Target ROAS. Switch to this once you have at least 30 conversions in the last 30 days. Below that floor, tROAS starves the campaign of volume.

When you do set a tROAS, set it near your actual recent performance, not near your aspiration. A target 40% above your current ROAS will simply throttle delivery. Move it in 10-15% increments and give each change two weeks.

Do not make structural changes during the four-to-six week learning period. Pausing asset groups, big ROAS target moves, or bidding strategy changes all reset learning. Most accounts that report poor PMax results modified the campaign mid-learning and never let it stabilise.

Creative assets: what to give it

Once you move beyond feed-only:

• Images: at least 5-8 per asset group, mixing product-on-white, lifestyle, and detail shots. Include 1:1, 4:5, and 1.91:1 ratios.
• Video: if you do not supply video, Google auto-generates it from your images and it is usually poor. Supply at least one real 15-30 second video per asset group.
• Headlines: 5-15, mixing product benefit, offer, and brand. Specific beats clever.
• Long headlines and descriptions: fill all available slots. Unfilled slots reduce the combinations the system can test.
• Logos: square and landscape versions.

Refresh cadence: PMax creative fatigues. Add three to five new assets every six to eight weeks. You do not need a full refresh; replacing the lowest-performing assets is enough. Asset performance scores (Best / Good / Low) appear in the Asset Group Performance report; replace Low-scored assets at the six-week mark.

Reading the reports honestly

This is where most PMax evaluation goes wrong. The reported numbers are real, but they answer a different question than the one you care about.

The harvesting signal

When reported ROAS climbs while blended customer acquisition cost and new-customer revenue stay flat, the campaign is harvesting existing demand rather than growing the store. That gap is the single most important diagnostic in PMax.

Track these together, monthly:

• PMax reported ROAS
• Blended ROAS (total revenue / total ad spend across all channels)
• New customer count and new customer revenue
• Blended customer acquisition cost

If PMax ROAS is up and blended ROAS is flat, you have a harvesting problem, and brand exclusions are usually the first fix.

Reconcile against Shopify

Compare PMax's reported conversion value against actual revenue in your Shopify P&L for the same orders and period. They should agree within 5%. Drift wider than 10% means either your conversion value configuration is broken (a tag firing twice, a theme update breaking the purchase event) or your conversion value is set to revenue when you meant contribution margin.

What the channel report is for

Channel-level reporting shows where budget went across Search, Display, YouTube, Discover, and Maps. Use it to justify better creative and feed work, not to judge whether PMax is growing the store. A high Display share early on often just means the algorithm has not found Shopping demand yet, which is exactly what feed-only launches avoid.

Common Performance Max mistakes on Shopify

• Not excluding your own brand. The default costs you incremental revenue and inflates your reported numbers. Fix this first.
• Launching with a weak feed. PMax spends most of its budget on Shopping. A thin feed caps everything downstream.
• Too many asset groups at launch. Splits budget, slows learning, produces noisy data.
• Changing things during learning. Four to six weeks means four to six weeks.
• Setting tROAS too early or too high. Below 30 conversions in 30 days, tROAS starves the campaign.
• Running PMax and Standard Shopping on the same products without exclusions. Internal competition inflates CPCs.
• Judging PMax on reported ROAS alone. Always alongside blended ROAS and new-customer metrics.
• Vague search themes. "Running shoes" tells the algorithm nothing it did not already know from your feed.
• Letting Google auto-generate video. Supply your own, even if it is simple.
• Never refreshing creative. Assets fatigue on a six-to-eight week cycle.

A sensible 90-day PMax rollout

Phase	Work
Weeks 0-2 (pre-launch)	Feed audit and cleanup. Custom labels set. Brand exclusion list built. Branded Search campaign live. Conversion tracking reconciled against Shopify.
Weeks 1-4	Launch feed-only PMax, one campaign, one asset group, brand excluded, Maximise Conversion Value with no target. Do not touch it.
Weeks 5-8	Add creative assets. Add first-party audience signals. Set tROAS near actual performance if 30+ conversions achieved. Add account-level negatives from search terms report.
Weeks 9-13	Evaluate on blended metrics, not reported ROAS. Consider a second campaign for margin segmentation. First creative refresh. Add first-party exclusions for a dedicated acquisition campaign.

When to hire

Setting up a PMax campaign takes an afternoon. Running one that grows a store rather than harvesting its existing demand requires feed work, incrementality thinking, and the discipline to leave it alone during learning. The hardest part is not the interface; it is knowing which reported number to ignore.

If your PMax ROAS looks good and your blended revenue has been flat for months, that gap is the whole problem and it does not fix itself. See how we run Google Ads for Shopify stores: feed first, brand excluded, judged on incremental new-customer revenue rather than platform-reported ROAS.

Frequently Asked Questions

Should I run Performance Max or Standard Shopping for my Shopify store?

Run both, segmented so they do not compete. PMax now has search terms reporting, brand exclusions, and account-level negatives, closing most of the old transparency gap. Standard Shopping still wins where you need explicit product-group bidding control or campaign priority tiers. Use listing group exclusions to keep them in separate lanes; without that you create internal competition that inflates CPCs without lifting volume.

What is the minimum budget for Performance Max?

PMax uses Smart Bidding exclusively and needs conversion volume to learn. Practically, you want at least 30 conversions in 30 days before switching to Target ROAS, and enough daily budget to accumulate that in a reasonable window. Under roughly $50-75/day, PMax struggles to exit the learning phase cleanly.

Do I need a separate brand Search campaign if I run PMax?

Yes. Exclude your brand from PMax and run a separate branded Search campaign with controlled bids. Branded traffic converts at a fraction of the CPC in Search compared to what PMax pays for the same clicks, and separating the two lets you judge PMax on incremental non-brand acquisition rather than harvested demand.

How long does Performance Max take to learn?

Roughly four to six weeks before performance stabilises. Structural changes during that window (pausing asset groups, significantly altering ROAS targets, changing bidding strategy) reset the learning period. Most accounts that report poor PMax results modified the campaign during learning.

What is feed-only Performance Max and should I use it?

Feed-only PMax runs Shopping-style ads from your Merchant Center feed without additional creative assets. It concentrates spend on Shopping placements where ecommerce conversion rates are strongest. Starting new campaigns feed-only for the first two to four weeks gives you a clean Shopping baseline before the algorithm starts allocating budget across Display and YouTube.