Performance Max in 2026: What Changed and What Actually Works
PMax replaced Smart Shopping and Local campaigns. Asset groups, search themes, and first-party data are now the levers. Here's what the best-performing PMax accounts are doing — and what to stop fighting.
Performance Max (PMax) is Google's fully automated campaign type that serves ads across all Google inventory simultaneously: Search, Shopping, Display, YouTube, Gmail, and Maps. It replaced Smart Shopping campaigns in 2022 and Local campaigns in 2023. By 2026, it accounts for the majority of Google Ads revenue for most e-commerce advertisers.
The fundamental shift PMax represents: you no longer control where, when, or to whom your ads show. You provide assets, signals, and conversion goals — Google's AI handles the rest. Most advertisers' initial instinct is to fight this. That's wrong. The right response is to learn how to give the algorithm better inputs.
How PMax Works in 2026
PMax campaigns run on a goal-based bidding system (Target ROAS or Target CPA). The algorithm distributes budget across all Google surfaces to hit your conversion goal. Within the campaign, you organise assets into Asset Groups — thematic collections of creative, headlines, descriptions, and (for e-commerce) product feeds.
Google reads your asset groups to understand what you're selling and who would be interested. It cross-references this with user intent signals across all its properties: Search queries, YouTube watch history, Gmail interactions, Maps searches.
Asset Groups — The Most Important Lever
Asset groups are the fundamental unit of PMax. Get these right and PMax is powerful. Get them wrong and you'll have a campaign that cannibalises budget across irrelevant products.
Best practices for asset groups in 2026
- One asset group per product category — don't mix running shoes and formal shoes in one group
- Separate high-margin and low-margin products — if you can't control budget allocation, separate them into different campaigns entirely
- Asset quality matters: Add 5 images, 5 logos, 5 videos, 5 headlines, 5 descriptions per group. More assets = more combinatorial testing = better results
- Video is mandatory at scale — PMax will create auto-generated video from your images if you don't provide video, and auto-generated video is almost always lower quality
Search Themes — Not Keywords, But Close
In 2024, Google added Search Themes to PMax — user-defined intent signals that guide where the algorithm serves on Search. These are not keywords (you can't add negative keywords at the theme level in the same way), but they're your strongest lever to influence search intent targeting.
In 2026, best practice for search themes:
- Add 5–7 search themes per asset group covering the primary intent (e.g., "buy running shoes online India," "men's training shoes under 3000")
- Use themes that match the actual language your target buyers use — pull from your Search Terms report in standard Search campaigns
- Review and update themes monthly based on what's converting
First-Party Data Is Now Critical
PMax's targeting quality improves dramatically with Customer Match data. Upload your customer list (hashed email + phone) to Google Ads and set it as an audience signal for your asset groups. Google uses this to:
- Build lookalikes of your best customers across all inventory types
- Suppress existing customers from acquisition asset groups
- Improve intent matching on Search
Accounts with Customer Match lists consistently report 20–35% better ROAS on PMax vs those without. Update your customer list monthly.
Protecting Your Brand: Search Brand Campaigns
PMax will bid on your brand keywords. If you don't have a dedicated brand Search campaign, PMax will capture brand searches — which inflate ROAS (brand searches convert easily) but make PMax's efficiency look artificially better. Always run a separate brand Search campaign with Exact Match keywords and use brand exclusions in PMax (available as a placement exclusion workaround in 2026).
What Actually Works vs. What to Stop
| Stop Doing | Start Doing |
|---|---|
| One asset group for all products | Separate asset groups by category + margin tier |
| No video assets (auto-gen) | Upload real product/brand videos |
| No Customer Match uploaded | Monthly Customer Match list updates |
| Evaluating PMax performance weekly | Give PMax 4–6 weeks to optimise before evaluating |
| Letting PMax bid on brand terms | Separate brand Search campaign + PMax exclusions |
| Generic product feed titles | Keyword-rich, specific product titles in feed |
How to Measure PMax Correctly
ROAS at campaign level is the right metric, but break it down by asset group to identify which product categories are performing. The Insights tab in PMax shows search categories, audience segments, and top-performing asset combinations. Review this weekly. Use it to improve feed titles, search themes, and creative.
Frequently Asked Questions
Should I use Target ROAS or Target CPA bidding in PMax?
For e-commerce with consistent AOV, use Target ROAS. For lead generation or variable order values, use Target CPA. Start with a conservative target (10–20% below your actual performance) and tighten it gradually over 2–3 weeks as the algorithm gathers data.
Can I use PMax for lead generation, not e-commerce?
Yes, but results are more variable. PMax for lead gen works best when you have high-quality conversion data (form submissions, calls, appointments) with clear intent signals. Without volume and quality conversion signals (100+ conversions/month), PMax may not outperform standard Search campaigns for lead gen.
How long does PMax take to exit the learning period?
PMax needs a learning period of approximately 6 weeks (not 2 like standard campaigns). During this period, avoid significant budget changes (stay within ±20% adjustments). Evaluate performance at weeks 4, 6, and 8 — not before. Premature changes restart the learning period.
Stop guessing. Start fixing.
Connect your Meta and Google Ads accounts to Pulse and get a full AI audit in minutes — free.
Try Pulse free →