Google Ads Performance Max: What Actually Works for eCommerce

Performance Max works for eCommerce when — and only when — you control its inputs: a clean product feed, asset groups split by margin or category, brand term exclusions, and a separate Manual CPC brand capture campaign running in parallel. Without those four guardrails, PMax will silently cannibalize your branded search, report a fictional ROAS, and hide what's actually driving new-customer revenue.

Does Performance Max actually work for eCommerce in 2026?

Yes — but the "PMax just works" narrative that Google pushes is misleading for most eCommerce brands. PMax is a black-box bid engine that optimizes toward whatever it can claim as a conversion. If you let it consume your brand terms, it will, because branded queries convert at 20–40× prospecting rates. You'll see a beautiful blended ROAS — and zero actual growth.

Across the accounts we audit, the single biggest PMax mistake is running it without brand exclusions. Fix that, and you'll see PMax ROAS drop on paper — while your actual new-customer acquisition improves. That's the trade worth making.

How should I structure Performance Max asset groups?

Split asset groups by product category or margin tier — not by audience signals. Google's audience signals are directional hints at best; the bidding algorithm largely ignores them once it has conversion data. What it does respect is the target ROAS you set per campaign, and it will optimize across all products in the asset group against that target.

The practical rule: products with similar unit economics go in the same asset group. High-margin hero products get one asset group with a lower tROAS (so PMax spends aggressively). Low-margin clearance or accessories get a separate asset group with a higher tROAS. Mixing a €20 accessory and a €200 hero product in the same asset group is how you get a ROAS number that means nothing.

A typical structure for an eCommerce brand:

• PMax — Hero Products (1 asset group, tROAS 2.5–3.5×, daily budget 50%+ of PMax spend)
• PMax — Category Best Sellers (2–3 asset groups split by category, tROAS 3.5–5×)
• PMax — Full Catalog (1 asset group, tROAS 5×+, the long tail)

Should I exclude brand terms from Performance Max?

Yes, always. Request campaign-level brand exclusions from your Google rep or via the interface (available to most advertisers now). Then run a standalone Manual CPC Search campaign for your brand terms, where you control bids, match types, and ad copy.

Why this matters: brand search typically runs at 20–40× ROAS because the customer is already looking for you. When PMax consumes those queries, it reports that ROAS as its own — and the algorithm learns "brand queries = cheap conversions" and spends more there, less on true prospecting. You end up paying PMax margin to capture demand you already had.

What feed quality issues kill PMax performance?

Feed quality is the invisible lever. Google's algorithm can only be as good as the signals you give it. The five issues we fix most often:

1. Product titles that match the product page H1. Titles should lead with category + brand + key attributes (material, size, color), not marketing taglines. "Organic Cotton Oversized Hoodie — Brand Name" beats "The Softest Hoodie You'll Ever Own".
2. Description stuffed with the word "our". Google reads descriptions for query matching. Remove the brand voice and add specs, materials, use cases, fit.
3. Product_type and google_product_category both empty. These drive category-level bidding. At minimum, map every product to a Google taxonomy category.
4. Custom labels unused. custom_label_0 through 4 let you segment by margin, season, or velocity. Use at least one for margin tiering so you can bid differently.
5. Out-of-stock products still active. Set automatic rules to disable zero-inventory SKUs. Serving ads for unavailable products wastes spend and hurts landing page experience scores.

How much budget does Performance Max need to learn?

Rough rule: 50 conversions per asset group over a 30-day window. For most eCommerce brands that's €3–5K/month per PMax campaign minimum. Below that, the algorithm doesn't have enough signal to optimize meaningfully — it's making guesses and you're funding the exploration.

If you're under that threshold, consolidate. Run one PMax campaign with two asset groups (hero + everything else) instead of five sparse ones. Sparse asset groups look like optionality but cost you data density.

The weekly Performance Max audit we run on every client account

Takes 15 minutes. Catches 90% of the leaks.

• Insights → Asset report: check which creative assets are "Best" vs "Low". Replace Low assets before they drag down asset group performance.
• Insights → Search terms: scan for brand terms sneaking back in, competitor terms, or irrelevant long-tail queries. Add negatives at account level.
• Placements report: exclude low-quality Display placements (apps, mobile games, etc.) — can save 15–25% of waste.
• Product performance: identify low-velocity SKUs eating budget. Move to a separate asset group with a higher tROAS or exclude.
• tROAS drift: if CPA is drifting up, the tROAS is likely too aggressive for current demand. Loosen by 10–15% temporarily to let the algorithm breathe.

When does Performance Max not make sense?

Three scenarios where we'd steer a client away from PMax or severely limit it:

• New brand, no conversion history. PMax needs data to optimize. Start with Standard Shopping + Search, build 30+ days of conversion data, then introduce PMax.
• High-AOV, long consideration cycles. PMax optimizes for immediate conversions. If your AOV is €500+ and the decision takes weeks, PMax will misfire. Go Search + Display + remarketing with view-through measurement.
• Very small catalog (under 20 SKUs). PMax assumes catalog depth. For a 5-SKU brand, Standard Shopping with manual bidding often outperforms.

Want to know how your ads stack up? Get a free audit at audit.rebel.online.