Performance Max 2026: the complete guide

Performance Max drives 20 to 34% Search cannibalization in median e-commerce accounts 2025-2026 according to our sector panel: a significant share of PMax budget buys back traffic already earned through brand. Mastering PMax in 2026 starts with mastering its domino effect on the rest of the account.

Performance Max has become the default Google Ads campaign. On the SteerAds 2025-2026 sample, PMax now represents 42% of Google Ads budgets invested, up from 18% two years earlier. Google pushes this format aggressively, advertisers follow, and the gains can be spectacular: +18 to +28% conversions at constant budget on well-configured accounts.

The problem: PMax is a black box that punishes improvisation hard. A poorly segmented asset group, thin audience signals, no brand exclusions, a budget below the critical learning-phase threshold — and you burn €3,000/month buying back traffic you already had for free. This guide lays out the complete method for launching, running, and auditing PMax in 2026: 10 key points, a 90-day roadmap, 4 key questions to answer before scaling.

What is Performance Max and how does it differ from Search?

Performance Max (PMax) is a unified multi-channel campaign that serves simultaneously across Google's 8 inventories: Search, Shopping, YouTube, Display, Discover, Gmail, Maps, and partner inventory. Where a Search campaign only reaches the Search network, and a Shopping campaign only reaches product results, PMax lets Google's algorithm decide, click by click, which channel, format, and bid to serve a given user. Launched in November 2021, it officially replaced Smart Shopping and Discovery Ads in September 2022 — the forced migration left many advertisers unprepared.

Technically, PMax rests on three elements: an asset group (the equivalent of an ad group, see section 3), one or more audience signals (priors you give the algorithm, not strict targeting, see section 4), and an automatic bidding strategy (Max Conversions or Max Conversion Value, with or without Target CPA/ROAS). Google's algorithm aggregates these three inputs with its own predictive model and decides the rest — placements, bids, formats, serving order. Official documentation is on Google Ads Help.

What's changed concretely in 2026: Google has significantly improved brand safety controls (finally granular brand exclusions since March 2024), introduced Asset Group Insights (per-asset-group detail with conversion rates and winning combinations), opened the PMax API to more exportable fields, and enabled finer audience signals (granular customer match, combined in-market audiences). The PMax of 2022 is no longer the PMax of today — but most online literature is out of date.

The fundamental difference from a classic Search campaign: you no longer choose keywords, no longer set bids, no longer target audiences in the strict sense. You influence the algorithm through signals, creatives, and exclusions. That makes PMax both very powerful (rapid scaling) and very tricky (no direct lever to correct a bad placement). The rest of this guide is structured around that idea: how to influence without controlling.

Where does Performance Max win and where does it lose?

Before launching PMax, you need to honestly assess whether your context is a fit. In most cases, 28 to 38% of accounts would have a better ROAS without PMax — often because their catalog, volume, or tracking don't meet the prerequisites.

PMax wins big on:

Multi-product e-commerce with a clean Merchant Center feed, 50+ SKUs, and clear seasonality.
Budgets ≥ €3,000/month allowing 50+ conversions / 14 days (learning-phase exit condition).
Strong audience signals: a Customer Match base of 5,000+ contacts, loaded 30/90/180-day site audiences.
Tracking maturity: Enhanced Conversions active, data-driven attribution, accurate conversion value.
Recognized brands with significant direct/organic traffic — PMax amplifies existing awareness.

PMax often loses on:

B2B SaaS with long sales cycles (> 30 days) and offline conversions not uploaded to Google.
Budgets < €1,500/month: never enough volume to exit the learning phase, PMax stays stuck in exploration.
Complex attribution with multi-touch custom weighting: PMax forces Google's model, not yours.
Ultra-specialized niches where YouTube/Display placement relevance is close to zero.
Sensitive brands (financial services, healthcare) where Display placement brand safety is critical.

Put differently: PMax is an accelerator for already-healthy accounts — not a fix for fragile ones. If your account is under-optimized on tracking or structure, run through our Google Ads audit checklist first before considering PMax. Launching PMax on a broken account just amplifies the waste.

How do you structure your PMax asset groups?

The asset group is to PMax what the ad group is to Search: the unit of strategic coherence. An asset group bundles creatives (images, videos, headlines, descriptions, logos), a set of final URLs, an audience signal, and optionally a search theme. Unlike a Search ad group, a PMax asset group doesn't have its own bids — those stay at campaign level.

The optimal number: 3 to 5 asset groups per campaign. Fewer than 3: you're underusing the ability to segment creatives finely by audience. More than 7: the campaign dilutes, each asset group gets too little data to perform. On the accounts we monitor continuously, campaigns with 4 to 5 asset groups outperform single-asset-group campaigns by 11 to 18% in ROAS.

The 3 segmentation logics to choose from:

By theme / product line (e.g., fashion e-com: "Dresses", "Jackets", "Accessories") — the most common.
By audience / funnel ("New customers", "Remarketing", "VIP Customer Match") — powerful when the CRM base is rich.
By offer / promo ("Winter sale", "New arrivals", "Best sellers") — useful for seasonal campaigns.

Fatal mistake :

1 single asset group for 200 products. That's the default configuration when you let Google handle it — and the one you find in 52 to 64% of audited PMax accounts depending on vertical. Result: the algorithm can't differentiate performance by segment, all creatives are mixed, impossible to optimize. Median ROAS observed on these accounts: around 2.1x (1.8-2.5x by sector) vs 3.4x on well-segmented accounts.

Concrete example (fashion e-com, €8k/month budget): 4 asset groups — (1) Women / Dresses / 25-44, (2) Women / Accessories / 18-34, (3) Men / Shoes / 25-54, (4) VIP Customer Match / Cross-gender. Each asset group gets 15 dedicated images, 5 videos, 5 short headlines, 5 long ones, 3 logos, and a specific audience signal. Budget pooled at the campaign level, but insights by asset group. This granularity is what makes PMax actually steerable.

How do you use audience signals without blocking the algorithm?

This is the most misunderstood aspect of PMax. Audience signals are not targeting in the Search sense. They're Bayesian priors: you're telling the algorithm "start looking here, these users are promising," and it uses those signals to bootstrap its exploration — then moves beyond them. With quality signals, PMax finds your audience faster; with thin signals, the learning phase lasts 3 weeks instead of 7 days.

The 5 signal types to combine:

Customer Match — your existing customers uploaded as hashed emails. See the Google docs. Strongest signal.
Website visitors segmented by recency (30 / 90 / 180 days) and page visited.
Google in-market audiences (e.g., "Women's clothing", "CRM software") — 2 to 4 per asset group.
Affinity audiences broader, for seeding new markets.
Custom segments based on keywords or URLs visited on third-party sites.

The 10-segment minimum rule. For an asset group to cleanly exit the learning phase in 14 days, it needs at least 10 audience segments combined as signals. In our sector panel, campaigns with 10+ signals show +14 to +22% ROAS vs those with 3-5 signals, over the first 60 days, variable by creative quality. Don't overload either — 25 signals isn't better than 15, redundancy dilutes.

Reminder: PMax will ignore these signals if it finds better elsewhere. That's normal and intended. Your role is to supply an intelligent starting point — which is essentially what Smart Bidding does in the background on Search campaigns, with comparable mechanics.

How do you avoid the PMax over-attribution trap?

Open the Google Ads interface. PMax shows 340 conversions for the month, 4.8x ROAS. Looks great. Except a significant share of those conversions would have landed without PMax — they're simply re-touched by a YouTube or Display placement in the user journey, and PMax takes the credit. That's the over-attribution phenomenon, and it's massive: PMax over-attributes between 18% and 32% of its conversions depending on funnel depth.

The cause: PMax serves across 8 channels, several of them low-intent (Display, YouTube, Gmail). Those impressions count as "view-through conversions" or "engaged-view conversions" — conversions attributed to someone who saw the ad without clicking. In a data-driven model, these view-throughs carry little weight, but they still carry some. And the Google Ads report adds them as-is to the PMax total.

The reliable method: 4-week holdout test. Cut PMax in one representative geographic region (e.g., Texas) for 4 consecutive weeks. Let it run normally elsewhere. Measure the gap in total conversions (all campaigns combined) across both perimeters, normalized by Search impressions. The difference between "conversions attributed to PMax" and "real incremental conversions" is your over-attribution factor. Across 100+ tested accounts, observed median: PMax over-attributes by 20 to 28% (IQR depending on account maturity).

Lighter alternative: Google Ads Geo Experiments, which automate this type of test. Documentation on Google Ads Help. To go further on attribution and the right metrics, read our ROAS / CPA / CPC guide.

Why are brand exclusions mandatory?

By default, PMax bids on all relevant searches for your account — including searches for your own brand. But those brand clicks, you earn almost for free through organic SEO and your dedicated brand Search campaign. PMax gets credit for them with a fantastic ROAS (brand conversion rate is always high), but the incrementality is zero. Reason #1 PMax "performs poorly" on a mature account: it buys back brand traffic already earned.

Since March 2024, Google has finally made brand exclusions granular via the "Brand safety" menu at campaign level. Before that, you had to go through Google support. Now it's a checkbox — and yet, in practice, 42 to 52% of PMax accounts still have no brand exclusions activated. Estimated average loss: €780/month per account.

The 3 types of exclusions to configure:

Brand exclusion list — your own brand + variants (typos), partner brands not to cannibalize. Procedure on Google support.
URL exclusions — blog pages, legal pages, support pages where you don't want to serve.
Product exclusions (feed) — out-of-stock SKUs, low-margin products, end-of-life products that PMax would keep pushing.

Key insight :

after activating brand exclusions, you typically see an apparent 18 to 25% drop in PMax ROAS. That's normal, and it's good news: the previously displayed ROAS was inflated by brand. What matters is the overall account ROAS, which remains stable or rises 4 to 7% when you reallocate the freed budget toward non-brand segments.

How do you work around the black-box reporting with scripts and API?

The native Google Ads interface for PMax is deliberately thin. No precise placement report on YouTube, no clear Search / Shopping / Display split within a single campaign, no match types on Search queries absorbed by PMax, no average CPC by channel. Google argues this opacity "protects the algorithm" — the reality is that it also prevents the advertiser from detecting wasteful placements.

Fortunately, 3 workarounds exist to recover this data, all proven on the accounts we manage.

Google Ads Scripts — notably the PMax Insights Script (free, open source, extensively documented). It runs in JavaScript within Google Ads, extracts details per asset group, per audience signal, per Display/YouTube placement, and generates an actionable Google Sheet. 15-minute installation. Official documentation on developers.google.com.
Google Ads API — for advanced accounts, access to fields not exposed in UI (e.g., asset_group_product_group_view, campaign_search_term_insight). Reference on developers.google.com/google-ads/api. Lets you feed a custom dashboard (Looker Studio / Data Studio) with views the UI doesn't allow.
Custom Looker Studio / Data Studio — native Google Ads connector + calculated transformations. You manually reconstruct metrics not accessible otherwise: ROAS per asset group, CPA per signal, ratio of Search vs Display impressions by estimation.

That's exactly what our Auto-optimization module does in the background: it calls the Google Ads API hourly, rebuilds opaque metrics, and alerts you as soon as an asset group drifts. For more on clean tracking, see also our conversion tracking guide.

How do you measure Search × PMax cannibalization?

PMax also serves on the Search network. If you have classic Search campaigns running in parallel (brand or non-brand), PMax can compete with them on the same queries. Result: you bid against yourself, CPCs climb, both campaigns steal conversions from each other. That's Search × PMax cannibalization, and it's systemic on poorly segmented accounts.

Google's implicit rule: when a query is eligible for both Search and PMax, Search wins if its ads are equally or more relevant. In practice, PMax very often takes the lead — because its audience signals make it more precise. Your non-brand Search campaigns quietly empty out without apparent explanation.

Holdout method to measure:

Cut PMax for 2 weeks in one region (a representative, isolatable market is often a good choice).
Let Search run normally everywhere.
Compare Search metrics (impressions, conversions, CPA) on the test zone vs control zones.
If your brand Search conversions rise by more than 12% in the test zone, cannibalization is confirmed.
If brand Search CPA drops by more than 8%, you were bidding against yourself.

Depending on vertical, one in three PMax campaigns significantly cannibalizes Search (28 to 38% of observed accounts). The standard fix: activate PMax brand exclusions (section 6), isolate your brand Search campaigns with exact-match keywords, and retest 4 weeks later. Median gain on brand CPA: -8 to -14%.

Warning :

never cut PMax across the entire account at once to "test." You'd lose 40-60% of your total volume and the restart triggers a full learning phase (14 days). The geographic holdout is the only clean method.

What minimum budget do you need to launch PMax without burning cash?

Google officially recommends 50 conversions over 14 days to exit the learning phase. That's the absolute minimum. Below that, PMax stays stuck in exploration — it allocates budget erratically, YouTube and Display placements balloon, CPA doubles for 3 to 4 weeks before stabilizing (or not). Underestimating this threshold is the 2nd cause of PMax failure on the accounts we audit, after the absence of brand exclusions.

Minimum test budget calculation:

Target CPA €30 → 50 conv × €30 × 1.15 (learning buffer) ≈ €1,725 over 14 days
Target CPA €50 → 50 conv × €50 × 1.15 ≈ €2,875 over 14 days
Target CPA €100 → 50 conv × €100 × 1.15 ≈ €5,750 over 14 days
Target CPA €200 → PMax probably unsuitable, test budget > €11,000/14 days rarely worth testing

Multiply by 2 for stable monthly budget post-learning phase. In other words: don't launch PMax below €3,000/month if your CPA exceeds €50. Below that threshold, you pay 2-3 months of unprofitability with no guarantee of recovery.

To understand Quality Score's and CPC's impact on this calculation — especially in competitive markets — see our Quality Score guide. When CPC climbs to €4-5, minimum PMax budget climbs proportionally.

What 90-day roadmap should you follow?

Here's the exact sequence we apply to launch PMax on a fresh account or to reset a drifting existing PMax. 90 days, 4 phases, precise actions by window.

If you don't feel equipped to run this sequence in-house, launch a free SteerAds audit: it automatically detects whether your account is ready for PMax, what minimum budget to target, and which audience signals are missing. For complex accounts (MCC multi-market, legacy ERP, need for hands-on support), reach us via the contact form. For accounts already in PMax looking to reduce CPA before scaling, follow up with our 10 levers in the complete CPA guide.

Sources

Official sources consulted for this guide:

FAQ

Should you keep your Search campaigns when launching PMax?

Yes, absolutely, and especially the brand ones. Performance Max will naturally try to buy back brand traffic because the ROAS looks excellent there — it's a mirage, that traffic you already earn for free through organic results. Keep a dedicated brand Search campaign, activate brand exclusions on PMax (see section 6), and use PMax for non-brand and cross-channel. On our internal SteerAds benchmark of 2,000+ audited accounts in 2025-2026, cutting brand Search in favor of PMax drove CPA up 23% on average in the first 30 days.

Does PMax replace Google Shopping?

PMax replaced Smart Shopping and Discovery in September 2022, but not Standard Shopping. The difference: Standard Shopping still offers granular per-product reports, negatives, and bid control. PMax hides all that in a unified black box. For an e-commerce business with a clean feed and a budget of €3,000/month or more, PMax outperforms Standard Shopping by 12 to 18% in ROAS according to our internal SteerAds benchmark. For niches, catalogs under 50 SKUs, or budgets below €1,500/month, Standard Shopping remains more controllable and often more profitable at the margin.

How long before PMax becomes profitable?

Expect 4 to 8 weeks for PMax to reach cruising speed. The first 14 days are a learning phase: performance is deliberately erratic, Google explores placements and audiences heavily. The theoretical minimum for profitability requires 50 conversions over 14 days — below that, the algorithm doesn't have enough signal. On our internal SteerAds benchmark, the median observed time-to-profitability is 42 days, with a high standard deviation depending on the quality of audience signals provided at launch. Plan for €2,500 to €4,000 of initial test budget without touching the settings.

How do you tell if PMax is cannibalizing your brand?

Reliable method: a 2-week holdout test on one geographic region. Cut PMax in a representative area (e.g., one region), let it run everywhere else. Measure the change in conversions and CPA on pure brand Search across both perimeters. If cutting PMax lifts your brand Search conversions by more than 12%, you have a cannibalization problem. Complement this with: activate brand exclusions in PMax (see section 6), use the PMax Insights Script, and compare the New Customer Acquisition column over the last 30 days — it should represent at least 40% of PMax conversions, otherwise you're buying back your own audience.