North American Media Experts
Performance Max can drain budget fast. Use this 7-step 2026 framework to structure asset groups, search themes, brand exclusions, and bidding for real PMax control.
9 min read
Performance Max is the most powerful — and most opaque — campaign type in Google Ads. One PMax campaign reaches across Search, Shopping, YouTube, Display, Discover, Gmail, and Maps from a single ad group. That sounds great until the algorithm starts spending 60% of your budget on a low-intent placement, cannibalizes your branded search, and you have no visible levers to pull. The fix is not "more automation." The fix is structure. In this guide, you'll learn how to structure a Performance Max campaign in 2026 using a seven-step framework that gives you real control over where the money goes, what creative shows, and which conversions count. Use it as a setup checklist for new accounts and as a teardown blueprint for accounts that have drifted.
Three updates shipped between late 2024 and early 2026 changed how seriously you need to take account structure. Google rolled out campaign-level negative keywords (you can now apply up to 10,000 per campaign without filing a support ticket), raised the search themes limit from 25 to 50 per asset group, and exposed channel-level reporting so you can finally see how spend is splitting across Search, Display, YouTube, Discover, and Maps. The implication is simple: PMax is no longer a black box, but only if you build the campaign in a way that lets the new controls bite.
Accounts that ignore structure tend to share the same symptoms — flat ROAS for months, runaway spend on irrelevant queries, branded conversions misattributed to PMax, and asset groups that all converge on the same hero image and one headline. Accounts that follow a disciplined seven-step structure regularly hit 3x to 5x ROAS on non-branded traffic, with budget distributed across channels in proportion to actual incremental contribution. Below is the exact sequence we use when we set up Performance Max for clients at North American Media Experts.
The single biggest structural mistake in PMax is putting two conflicting goals inside one campaign. Lead generation and e-commerce sales optimize against fundamentally different signals — average order value vs. cost per lead, days-to-conversion vs. instant purchase, on-page form events vs. checkout events. When the algorithm is told to optimize for both, it picks the cheaper signal and starves the more valuable one.
Run one Performance Max campaign per distinct business goal. For e-commerce, that might be one campaign for full-catalog Shopping plus three campaigns segmented by margin tier — premium, standard, clearance. For a B2B SaaS company, it might be one campaign for free-trial signups and one for enterprise demo requests. The number is less important than the principle: each campaign needs a single, unambiguous conversion definition and a budget large enough to escape learning purgatory. As a working benchmark, plan for at least 30 conversions per month per campaign before you expect stable performance.
If you're staring at an account where one PMax campaign is trying to do everything, pause before you split it. Pull a baseline by running our 10-step Google Ads audit checklist first so you know what "before" looks like.
Performance Max is a smart-bidding campaign at its core. Its targeting is only as good as the conversion data it eats. Two changes pay for themselves before you touch a single asset.
First, audit your conversion goals. Demote micro-conversions like page views, scroll depth, and add-to-cart to secondary status so they inform the model without dictating bids. Promote primary purchase, lead form completion, or qualified-lead webhook events to primary goals. A common pattern: a SaaS account had "newsletter signup" set as a primary goal worth $5, while a $1,200 demo request sat as secondary. Once we swapped the priorities, CPA on demo requests fell roughly 38% over six weeks because the algorithm finally knew what mattered.
Second, push value, not just count. If you sell products with variable margin, send back gross profit rather than the order total. If you sell leads to a sales team, fire enhanced conversions with the lead score once your CRM scores it — Google ingests offline conversion uploads natively. Value-based bidding strategies (Max Conversion Value with a tROAS target) only outperform Max Conversions when the algorithm has real value variance to work with.
Inside each PMax campaign, asset groups are the unit of organization. The 2024 best practice of segmenting asset groups by audience signal is dead. Google's own guidance for 2026 is unambiguous: one theme, one asset group, where a theme is a product line, a service category, or a value proposition.
For an e-commerce home goods retailer, that might be four asset groups: indoor furniture, outdoor furniture, lighting, and rugs. For a multi-location dental group, themes might be cosmetic dentistry, family dentistry, emergency care, and orthodontics. Each gets dedicated headlines, descriptions, images, video, and landing page — assets that genuinely match that theme rather than recycled brand-level creative.
Two rules tighten this further. Do not duplicate asset groups by audience signal — a "high-intent" version and a "lookalike" version of the same theme split your data and undermine learning. Do not split asset groups below 30 conversions per month — under that threshold, the algorithm cannot distinguish the groups and you'll see erratic delivery. When in doubt, consolidate. If you need true audience containment, look beyond PMax — start with our guide to audience targeting and segmentation.
Search themes are the closest thing PMax offers to a steering wheel. In 2026, you get up to 50 search themes per asset group, and Google treats them with the same prioritization as phrase-match and broad-match keywords in a Search campaign. That makes them a strong "hint" — but only if the themes are specific.
The themes that work look like real long-tail queries: "corporate gift baskets toronto", "emergency root canal sunday", "linkedin retargeting agency b2b saas". The themes that fail are vague brand-speak like "premium service" or "high quality". Pull your top-converting search queries from the past 90 days, pull rising queries from Google Search Console, and add long-tail variants that match your asset group theme. Aim for 15 to 30 themes per asset group — enough to direct the algorithm, not so many that the signal becomes noise.
Search themes are particularly valuable when you are migrating away from Dynamic Search Ads. If you're working through that transition right now, pair this section with our DSA to AI Max migration playbook so search themes inherit the right query coverage.
The fastest way to inflate Performance Max ROAS on paper is to let it eat your branded search. The fastest way to actually waste money is the same move. Brand exclusions are a non-negotiable structural step.
Add your brand and brand variants to campaign-level brand exclusions under Campaign Settings → Brand Exclusions. The list supports up to 1,000 entries per account; cover misspellings, abbreviations, and competitor co-occurrence terms. Layer this with campaign-level negative keywords — a control Google rolled out fully in 2025 and expanded to 10,000 negatives per campaign in early 2025. Use the negative list for irrelevant intent (free, jobs, careers, cheap), low-quality geographies, and direct competitors you don't want to spar with.
If you run a dedicated branded Search campaign, brand exclusions in PMax also force the algorithm to find incremental conversions rather than steal credit from the work your branded campaign already does. Set up an account-wide negative list under Tools → Shared Library so the same exclusions apply consistently across every PMax campaign you build.
Performance Max assembles ads on the fly across seven inventory types, which means the campaign needs enough raw material to build every format. The asset minimums that consistently outperform thin asset groups in 2026:
Creative quality matters as much as quantity. Asset groups that mix proven direct-response patterns with on-brand storytelling consistently outperform asset groups that only push promotions. For a deeper dive into what's working creatively this year, see our 2026 creative strategy trends breakdown and our broader creative services overview.
Performance Max needs roughly four weeks to learn a new structure. Optimizing too aggressively before then thrashes the model. Optimizing too rarely after then leaves money on the table. Our standard cadence:
Week 1. Set a tCPA or tROAS target that is 30% looser than your eventual goal. This buys the algorithm room to find conversions. Don't touch budgets, don't pause asset groups, and don't add new search themes mid-week.
Week 2. Review channel distribution in the new channel-level report. If 70%+ of spend is landing on Search, the algorithm is finding intent — good. If 70%+ is on Display, your search themes are probably too vague — revisit Step 4. Check asset ratings (Low, Good, Best) and replace any Low-rated assets.
Week 3. Tighten the tCPA or tROAS by about 10%. Pull the search terms insights report (expanded in 2026) and add any irrelevant queries to your campaign-level negatives. Promote the best-performing asset variants and retire the worst.
Week 4. Evaluate split opportunities. If a single asset group is generating 50+ conversions per month and contains two clearly distinct themes, split it. If two asset groups together are generating under 30 conversions per month, merge them. Re-baseline against the metrics you set in Step 1.
Repeat the four-week cycle continuously. Every meaningful change — new product line, new geographic market, new promotion — restarts the learning clock for that asset group, not the whole campaign.
Even with a clean framework, four mistakes show up in nearly every audit we run.
Mistake 1: One mega-campaign for everything. Multiple goals, dozens of asset groups, a single budget. The algorithm cannot distinguish optimization signals. Split by goal first, then by theme.
Mistake 2: Treating audience signals as targeting. Audience signals in PMax are hints, not constraints. Adding a "high-value customer list" as a signal does not restrict the campaign to that list. If you need real audience containment, you may need a different channel altogether — start with our paid social vs. programmatic channel mix guide to pick the right approach.
Mistake 3: No video assets. YouTube can quietly consume 30 to 50% of PMax budget. Without your video, Google auto-generates one. Auto-generated videos hurt brand perception and view-through rates.
Mistake 4: No budget discipline. Performance Max will spend whatever you give it. Setting daily budgets based on intuition rather than capacity math is how accounts overspend on low-intent placements. Run the numbers using our 2026 Google Ads budget guide before you scale.
Performance Max is not a "set it and forget it" campaign. It's a high-leverage system that rewards advertisers who give it clean inputs, clear signals, and the right guardrails. The seven steps above — one campaign per goal, clean conversion data, themed asset groups, specific search themes, strict brand exclusions, deep creative libraries, and a disciplined four-week optimization rhythm — turn PMax from a black box into a budget you can steer.
If you'd rather have a team that runs this framework for you across Google, Meta, programmatic, and CTV, our paid media specialists build and operate Performance Max accounts every day for brands across North America. Request a custom quote and we'll show you exactly how this structure would apply to your account, your conversion data, and your 2026 targets.