Performance Max Campaign Structure for Ecommerce: The 2025 Guide
Performance Max has become unavoidable for ecommerce Google Ads accounts, but most advertisers deploy it wrong: one single asset group, zero audience signals, and a shared budget with Search campaigns. The result: disappointing ROAS and zero visibility into where conversions actually come from. At MS4D, PMax structure is one of the first things we audit — and it's consistently where the fastest wins are hiding. This guide explains how to properly structure your Performance Max campaigns in 2025 to maximize ROAS, avoid the most common mistakes, and regain control over what Google does with your budget.
What is a Performance Max asset group?
An asset group is the basic building block of a Performance Max campaign. It bundles all your creative assets — images, videos, headlines, descriptions, logos — that Google will automatically combine to serve ads across its entire inventory: Search, Shopping, Display, YouTube, Discover, and Gmail.
Unlike Search campaigns where you control keywords and ad groups, Performance Max works by giving Google's algorithm a set of creatives and audience signals, then letting it optimize delivery autonomously.
The key distinction: an asset group is not a product group. It's a set of creatives linked to an audience signal and, optionally, a listing group (product filter). A single asset group can therefore serve ads for thousands of products if you don't configure listing groups.
Understanding this definition matters because most PMax structuring mistakes stem from confusing asset groups with product segmentation. We'll cover best practices in detail below.
Why PMax structure directly impacts your ROAS
Your Performance Max campaign structure determines two fundamental things: the algorithm's ability to optimize, and your ability to analyze performance.
**Impact on the algorithm.** Google optimizes PMax at the campaign level, not the asset group level. If you put all your products in one asset group with one audience signal, Google will chase overall conversions without distinguishing your high-margin products from your loss leaders. You end up subsidizing growth for your least profitable products.
**Impact on analysis.** With a single asset group, there's no way to know whether your conversions came from Display creatives, Shopping listings, or PMax-generated Search ads. You're flying blind.
**Search vs PMax routing post-Q4 2024.** Since late 2024, Google changed how traffic is routed between Search and PMax campaigns: it's now Ad Rank that decides which campaign wins the auction, not PMax priority. A Search campaign with a strong Quality Score can therefore capture traffic at the expense of PMax. Consequence: don't delete your Search campaigns assuming PMax "covers everything." Both need to coexist and complement each other.
Recommended structure: 1 asset group per product family
Our recommendation for ecommerce accounts: **1 asset group per product family that's homogeneous in terms of margin and marketing message**.
Concrete example for a fashion account: - Asset group "Winter Coats" — listing group: Coats category, lifestyle fall/winter creatives, audience signal: coat buyers + remarketing visitors to the coats section - Asset group "Running Sneakers" — listing group: Running Shoes category, sport/performance creatives, audience signal: sport affinity + sneaker buyers - Asset group "Accessories <€50" — listing group: price < €50, impulse-buy creatives, audience signal: broad
This segmentation enables three things: creatives that match the products being served, relevant audience signals per family, and readable performance data by product universe.
**How many asset groups maximum?** There's no technical limit, but beyond 5-6 asset groups per campaign, the algorithm lacks sufficient data per segment to optimize properly. Fewer well-fed asset groups outperform over-segmented structures where each segment lacks the conversion volume needed for learning.
Audience signals: guiding the algorithm without constraining it
Audience signals in PMax are not targeting settings. They're information you give the algorithm about the type of people likely to convert — it can go beyond them if it detects other opportunities.
**Priority signals (in order):** 1. Remarketing lists (product page visitors, cart abandoners, past buyers) — strongest signal, based on real behavior on your site 2. Customer Match (CRM customer list) — excellent signal for identifying high-LTV buyer profiles 3. Custom audiences (based on search intent: keywords typed on Google) — highly effective for categories with strong commercial intent 4. Affinity and in-market audiences — broader signal, useful for new categories
**What to avoid:** adding overly broad audiences (e.g., "all EU users") as your only signal. You're giving the algorithm an empty signal, which slows learning and dilutes budget.
**PMax generates an average 15-20% additional Search volume** compared to a Search-only campaign covering the same queries — that's the real PMax gain: incremental queries your Search keywords don't capture.
Migrating from Maximize Conversions to Maximize Conversion Value: the critical timing
A frequent mistake on PMax accounts: switching too quickly from "Maximize conversions" to "Maximize conversion value" (with or without a target ROAS).
The rule: **wait a minimum of 30 days of stable data before migrating**, and ensure you have at least 30 conversions recorded within the campaign's attribution window.
Why? The PMax algorithm needs a historical data baseline to calculate the expected value of a click before switching to value optimization. Without this base, it will bid randomly, the campaign will enter an extended learning phase, and your ROAS will drop.
**Recommended migration protocol:** 1. Launch with Maximize Conversions (no tROAS) for 30 days 2. Check: 30+ conversions, stable ROAS over the last 14 days 3. Switch to Maximize Conversion Value (no tROAS target) for 2 weeks 4. Add a tROAS target at 80% of your currently observed ROAS 5. Raise the tROAS target progressively in 10% increments every 2 weeks
Rushing this migration is one of the most common causes of PMax performance collapse on the accounts we audit.
The 3 PMax structure mistakes that kill ROAS
After auditing dozens of PMax accounts, here are the 3 most common structural errors:
**Mistake 1: Everything in one asset group.** A single "All products" asset group with generic creatives and no precise audience signal. The algorithm can't optimize finely, creatives don't match the products being served, and you can't analyze performance by family. Fix: segment into 3-6 asset groups by product family/margin.
**Mistake 2: Deleting Search campaigns.** Many advertisers believe PMax replaces Search. Since Q4 2024, this is no longer true. Keep your Search campaigns on high-intent keywords (branded, competitor brands, exact product queries). PMax handles incremental volume; Search secures strategic queries.
**Mistake 3: Activating tROAS too early or too aggressively.** A target ROAS of 800% on an account with an actual ROAS of 400% forces the algorithm to cut the vast majority of bids. The campaign enters "Restricted by bidding strategy" status and stops spending. Rule: your tROAS target should never exceed 120% of your observed ROAS in the initial phase.
Listing groups: segmenting products without overcomplicating structure
Listing groups in PMax work like product filters in standard Shopping campaigns. They let you include or exclude specific products in each asset group.
**Most useful attributes for segmentation:** - Product category (Google taxonomy label) - Brand - Custom labels (created in your Merchant Center feed) - Product ID (to isolate specific bestsellers) - Price (to separate product tiers)
**Recommended strategy:** use custom labels to create 3-4 profitability tiers in your feed (e.g., label "high-margin", "mid-margin", "clearance"), then build your asset groups around these labels. This lets you apply different bidding strategies based on actual product profitability, not just category.
**What to avoid:** segmenting at the individual product level. Beyond the management complexity, each overly granular listing group lacks enough data to feed the algorithm — especially on accounts with fewer than 100 monthly conversions.
Checklist: launching a PMax ecommerce campaign the right way
Before publishing your next Performance Max campaign, verify these 8 points:
1. **Complete, error-free Merchant Center feed**: titles, images, prices, availability all up to date 2. **Asset groups segmented by product family**: minimum 3 distinct asset groups for catalogs with 100+ products 3. **Audience signals configured**: at minimum, site remarketing + Customer Match 4. **Complete creative assets**: 15+ images, 5 logos, 5 videos (or auto-generated video enabled), 15 headlines, 4 descriptions 5. **Appropriate bidding strategy**: Maximize Conversions (no tROAS) for the first 30 days 6. **Isolated budget**: don't mix PMax and Search in a shared portfolio budget 7. **Active ad extensions**: sitelinks, callouts, structured snippets configured at the account level 8. **Exclusions configured**: campaign-level negative keywords + audience exclusion lists if needed (e.g., exclude recent buyers if your goal is new customer acquisition)
Key Takeaways
- 1 asset group per homogeneous product family by margin — never everything in one asset group
- Since Q4 2024, Search vs PMax routing is determined by Ad Rank: keep Search campaigns on strategic queries
- Wait 30 days and 30 conversions before switching from Maximize Conversions to Maximize Conversion Value
- Audience signals are not targeting settings: start with remarketing and Customer Match
- A tROAS target should never exceed 120% of your actually observed ROAS in the initial phase
This article is based on episode 0 of the podcast
Listen to the full version with Alexia and Maxence to dive even deeper.