Performance Max in 2026: Why Your Campaigns Are Underperforming (And the 10 Levers Most Advertisers Ignore)

A client came to us last quarter with a familiar story. Their Performance Max campaign was reporting a 6.2x ROAS. The CMO loved the dashboard. Quarterly spend was $180,000.

We pulled three numbers nobody had been looking at.

First, the share of total PMax conversions coming from branded search queries: 71 percent.

Second, the percentage of impressions on YouTube and Display in placements with under 60 percent viewability: 44 percent.

Third, the new-customer share of conversions: 19 percent.

That campaign wasn't generating 6.2x ROAS on incremental revenue. It was claiming credit for branded search traffic that would have converted on a $0.30 CPC manual search campaign, then burning the rest of the budget on cheap, low-quality impressions across Google's network. Net incremental ROAS was closer to 1.4x. This is the most expensive and most common pattern we see in Performance Max accounts in 2026. The campaign type now drives 45 percent of all Google Ads conversions across the platform, and advertisers using it correctly are getting genuinely transformative results. Advertisers using it on autopilot — which is most of them — are quietly hemorrhaging budget while the dashboard tells them everything is fine.

This piece is the diagnostic framework we use to audit Performance Max campaigns. It surfaces the 10 specific levers that separate scaling accounts from bleeding ones.

The reality of PMax in 2026

Performance Max is no longer the black box it was when it launched in 2022. The 2024 and 2025 update cycles added meaningful controls — campaign-level brand exclusions, channel performance reporting, search themes, asset group segmentation reporting, age/gender exclusions, account-level negative keywords as a system, and a real Search Terms report that's actually useful. Recent platform updates have further expanded what's visible: the video asset limit per asset group has been raised to 15, and Microsoft's PMax variant now supports up to 50 search themes.

The campaign type has matured. The advertiser base hasn't. A large-scale Adalysis study covering 3,300 accounts found that when Search and PMax campaigns competed for the same query, the manual Search campaign consistently won on conversion rate. That's not a failure of Google's AI — it's a signal about what PMax does when left ungoverned. It defaults to the easiest path: claiming branded search traffic and cheap Display impressions, then reporting impressive aggregate numbers that obscure where the value actually came from.

The job in 2026 is to govern the algorithm hard enough that it spends your budget on incremental conversions instead of on conversions you would have gotten anyway. That requires pulling levers most advertisers don't know exist.

The 10 levers most advertisers ignore

1. Brand exclusions at the campaign level — not negatives. This is the single most leveraged change you can make in most accounts. PMax does not respect campaign-level negative keywords the way Search campaigns do. The fix is the dedicated Brand Exclusions feature (Campaign settings → Brand exclusions), which prevents PMax from serving on queries containing your brand. If you don't set this up, PMax will eat your branded search like a buffet and credit itself for revenue that would have happened anyway. Audit any account: if brand exclusions aren't set, that's the first move.

2. Account-level negative keyword lists. PMax doesn't accept campaign-level negatives natively, so they must live at the account level — which means they affect Search and Shopping too. Build the list deliberately: competitor brand terms you don't want to bid on, informational prefixes ("how to," "what is," "free"), irrelevant adjacent categories, and geographic terms outside your service area. Most accounts we audit have either no negative list or one with under 30 entries. A working list is usually 200 to 800 terms.

3. Asset group segmentation by intent, not SKU. The default mistake is one giant asset group with everything in it. PMax cannot match the right creative to the right context when creative is undifferentiated. Segment by audience intent (research vs. ready-to-buy), product category, price tier, or campaign objective. Each asset group teaches the algorithm a coherent story. Segmented asset groups consistently outperform monolithic ones in our accounts by 25 to 40 percent on conversion rate.

4. Search themes derived from your top Search keywords. Search themes are PMax's keyword-guidance lever — up to 25 per asset group on Google, 50 on Microsoft. Most advertisers either skip them or invent generic themes from the brief. The right move is to mine your existing Search campaigns for the keywords already converting at high quality, then translate those into search themes per asset group. You're handing the algorithm the playbook your manual campaigns already proved works.

5. Audience signals as a teaching tool, not a targeting filter. Audience signals in PMax are widely misunderstood. They don't restrict who sees your ads — they teach the algorithm what your high-value customers look like, then PMax finds similar people. Upload Customer Match lists of your top buyers (last 24 months), past converters, high-AOV cohorts, and active engagers. The algorithm uses these as the seed pattern. Skipping this step is a common reason new PMax campaigns take 6+ weeks to find their footing.

6. New Customer Acquisition mode with proper value weighting. If your business cares about growth, not just retention, configure PMax for new customer acquisition. Enable "Only bid for new customers" or set a higher value weight for net-new buyers (2-3x is standard). Set the new-customer conversion value to at least 30 percent of your average order value. This single switch reorients PMax away from cannibalizing existing customers — which it will absolutely do by default — and toward genuinely incremental revenue.

7. Manual video over auto-generated. Google will auto-generate video from your image and text assets if you don't provide one. This is convenient and consistently 25 to 40 percent worse than manually produced video on conversion rate. The platform now favours video-enabled asset groups in its bid landscape, which means you're paying a performance tax for relying on auto-generated content. Tools like Veo, Runway, and even iPhone-shot UGC give you better video output than Google's auto-generator at this point. Upload three to five real videos per asset group.

8. Asset volume — match Google's spec, don't just hit minimums. PMax wants 15 headlines, 5 descriptions, at least 10-20 images across landscape (1.91:1), square (1:1), and portrait (4:5) formats, and 3-5 videos per asset group. Most accounts ship 5 headlines, 2 descriptions, and 3 images. Limited asset variety means limited combinations means limited optimization headroom. The algorithm needs raw material. Treating asset volume as compliance instead of fuel is one of the quiet reasons accounts plateau early.

9. The change-cadence rule. Every meaningful change to PMax — new asset group, ROAS target shift, audience signal update, big budget change — partially resets the learning algorithm. Advertisers who tweak every 2-3 days create campaigns that never fully optimize. The rule we operate by: no major changes more often than every 7 days, no ROAS target changes greater than 25 percent at a time, and a minimum 14-day evaluation window before judging any new asset group. Most underperforming accounts we audit are over-optimized — managers tinkering with them daily.

10. Channel-level performance reporting and asset ratings. This is where PMax has changed most in 2026 and where most advertisers haven't caught up. The platform now shows performance breakdowns by channel (Search, Shopping, YouTube, Display, Gmail) and rates every individual asset as Low, Good, or Best. Pull these reports weekly. If 60 percent of your conversions are coming from Search and 40 percent of your spend is going to Display with no incrementality, you have a structural budget allocation problem you can't solve from inside the campaign — but you can solve it by restructuring asset groups or splitting a campaign. If half your assets are rated Low, replace them. The data is finally there. Use it.

A 60-minute PMax audit you can run today

If you're managing PMax in-house and want to know where you stand, this is the audit pass we run as a first move on any new client account. It takes about an hour.

Open the campaign and check, in order: Are brand exclusions set? Is the account-level negative list populated with at least 200 deliberate terms? How many asset groups exist, and are they segmented by intent or just dumped together? How many search themes per asset group, and are they derived from converting Search keywords or made up? Are audience signals populated with Customer Match lists of real buyers? Is New Customer Acquisition mode on, with appropriate value weighting? Are real videos uploaded, or is Google auto-generating them? Does asset volume hit Google's recommended ceiling, not just the minimum? When was the last major change, and is the campaign over- or under-edited? What's the channel split on conversions vs. spend, and do the asset ratings tell a coherent story? If three or more of those answers are wrong, you've found at least 30 percent of upside available without spending an extra rupee. Most accounts we audit have six or seven wrong. The total upside on a misconfigured PMax program — assuming spend stays flat — is typically a 40 to 80 percent lift in incremental ROAS within 90 days.

When PMax is actually the wrong tool

A piece of honesty most agencies skip: PMax is not the right campaign type for every account. It earns its 45 percent share of Google's conversion volume on accounts that fit a specific profile. It's wrong for you if your monthly spend is under roughly $1,500 (PMax needs volume to learn), if your conversion tracking is incomplete or inaccurate (garbage signals create garbage optimization), if you need precise keyword-level control for highly strategic terms, if your product is in a heavily regulated vertical with creative restrictions, or if your business depends on tight attribution to specific channels that PMax intentionally blends.

In those cases, structured manual Search and Shopping campaigns will outperform PMax for the foreseeable future. We've migrated several accounts off PMax this year and the results have been better than the dashboard previously suggested. The right campaign type is the one that fits your business, not the one Google is most aggressively pushing.

What this means for your account this quarter

The advertisers winning with Performance Max in 2026 are not the ones spending the most. They're the ones who treat the campaign type as a system they have to govern actively, not a magic box they hand budget to. Brand exclusions on. Asset groups segmented. Search themes derived from real data. Customer Match audiences populated. Real videos uploaded. New customer mode configured. Cadence disciplined. Channel and asset reports read weekly.

That's the difference between a 6.2x reported ROAS that's actually 1.4x incremental and a 4.1x reported ROAS that's a real 4.1x. The platforms reward operators who do the work. They reward operators who don't with reports that look great and businesses that quietly stagnate. If you'd rather have someone audit your PMax campaigns, restructure the asset groups, build the negative lists, and ship the 90-day fix — that's the work we do at Praxxii Global. Across the accounts we manage, the average PMax ROAS lift from a structured restructure has been 47 percent in the first 60 days. That's not because we have a secret. It's because the levers exist and most accounts aren't pulling them. Run the 60-minute audit above. If you find more than three wrong answers, you have a problem worth fixing this month.