Performance Max in 2026: when it works, when it wastes budget

Performance Max (PMax) is a single campaign type that serves across Search, Shopping, YouTube, Display, Discover, Gmail and Maps using automated bidding, automated targeting from your audience signals and conversion data, and automated creative assembly from the assets you supply. You give it goals, budget, assets and signals; it decides where and to whom to show ads.

That trade is the whole story. You get reach and machine-driven optimization you could not run manually. You give up keyword-level control, placement control and a lot of visibility. Whether the trade is good depends on what you feed it.

Use it when most of these are true:

• Conversion volume is healthy. 30 or more conversions a month in the campaign, ideally many more. Below that the system is guessing.
• Conversion tracking is clean and complete. Purchase values for ecommerce, qualified-lead values or offline conversion import for lead generation, enhanced conversions and consent mode in place. PMax optimizes hard toward whatever you count, so counting the wrong thing is expensive.
• You have a product feed (Google Merchant Center) with good titles, images, prices and availability. For ecommerce, feed-based PMax is the main reason the campaign type exists.
• You can supply real creative. At least five to ten images in multiple ratios, several videos (even simple ones), and headlines and descriptions written for the offer, not placeholders.
• The goal is incremental customers. New-customer acquisition for ecommerce, or lead generation in a category where you already have Search covered and want more volume.
• Budget is at least five to ten times your target cost per acquisition per day.

In these conditions, US ecommerce advertisers commonly see PMax deliver volume beyond Search at a comparable or slightly higher cost per acquisition, with Shopping inventory doing most of the work.

Be skeptical when:

• Conversions are under about 15 a month, or the conversion action is a soft event like a page view.
• You have no feed and no video, so the campaign leans on Display and Discover placements with low intent.
• Your brand is well known and searched for often; without exclusions, PMax will take the brand clicks you would have won for pennies and report them as wins.
• The sales cycle is long and offline, and you have not set up offline conversion import; the campaign optimizes toward form fills, including junk ones.
• You are in a category with strict compliance (healthcare under HIPAA, financial services, legal) where creative and landing page variation needs review before going live. Automated asset assembly can produce combinations you did not approve.
• You need geographic or dayparting precision for a service business; PMax honors location targeting but gives you less control than Search.

In these situations, a well-structured Search campaign, Standard Shopping or a Demand Gen campaign for video and discovery usually beats it.

1. Keep Search for known winners. Exact and phrase match campaigns for your proven keywords. Google gives eligible Search keywords priority over PMax for the same query, so this protects your core.
2. Exclude brand. Use brand exclusion lists at the campaign level so PMax must find new demand. If you want brand coverage, run it in a separate Search campaign where you can see it.
3. Add negative keywords. Campaign-level negatives are available; use them for the obvious waste (jobs, free, DIY, competitors you do not want).
4. Set a new-customer acquisition goal when the objective is growth, with a customer match list so the system can tell new from existing. Choose the "bid higher for new customers" or "new customers only" mode deliberately.
5. Structure asset groups by product category or offer, not by audience. One asset group per theme with tailored headlines, images and landing pages. Three to eight asset groups is a reasonable range for most accounts.
6. Use listing groups for feed control. Exclude low-margin or out-of-stock products, and separate high-margin lines into their own campaign if they need a different ROAS target.
7. Start with Maximize Conversions or Maximize Conversion Value without a target for the first two to four weeks, then set a target CPA or ROAS near what the campaign achieved, and tighten in steps of 10 to 15 percent.
8. Give it audience signals, not targeting. Customer lists, converters, high-intent in-market segments. Signals steer the start; they do not limit reach.
9. Turn off final URL expansion or add URL exclusions if you do not want the campaign sending traffic to blog posts, careers pages or policy pages.
10. Review automatically created assets and disable them if the generated text does not meet your brand or compliance standards.

• Channel performance report: see spend and conversions by Search, Shopping, YouTube, Display and so on. If Display or Video is taking a large share of spend with few conversions, revisit creative and consider exclusions.
• Search terms insights: categories of queries the campaign served on. Use it to find negatives and to spot brand leakage.
• Asset group and asset performance: which images, videos and headlines earn "Best" or "Low" ratings. Replace Low assets every two to four weeks.
• Listing groups: product-level performance for feed campaigns; prune and promote accordingly.
• Incrementality check: compare total account conversions and cost for the four weeks before and after launch. If PMax conversions rose while Search and brand conversions fell by a similar amount, you shifted credit rather than created sales.

• Launching with thin assets and letting the system assemble generic ads from your logo and two product photos.
• Letting it run on brand and reporting the inflated ROAS as success.
• Changing budgets and targets every few days. Give it one to two weeks between changes.
• One asset group for the whole business. Messaging gets generic and landing pages do not match.
• No new-customer setting on an acquisition goal, so the campaign retargets your existing buyers.
• Judging it by platform ROAS instead of contribution margin and total new customers.
• Running it in low-volume lead gen where Search would convert more cheaply.

1. Channel split: where did the money go, and did each channel earn it?
2. Search terms and brand leakage: new negatives and exclusions.
3. Asset ratings: replace Low, add variants of Best.
4. Listing group performance: pause losers, isolate winners.
5. Target adjustment: one change, documented.
6. Incremental view: total account conversions and margin, not campaign ROAS.

We decide whether Performance Max belongs in an account based on signal quality and goal, not on the platform's recommendation. In discovery we audit conversion tracking, feed quality, creative inventory and the overlap between existing Search and brand demand. In planning we define the role PMax will play (incremental Shopping volume, new-customer acquisition, or none yet), the exclusions that protect the core and the measurement that will judge incrementality. In build we structure asset groups by offer, produce the creative assets the campaign needs, implement brand exclusions, negatives, customer lists and URL controls, and connect offline conversions where sales happen off-site. In improve we run the monthly review above and report on total new customers and margin alongside platform metrics. The paid media service is at /services/growth/performance-marketing; for feed and store work see /services/web/shopify-and-ecommerce.

Our Performance Growth Sprint is the representative engagement: a paid-media reset built around measurement first, structured creative testing, and commercial decisions made on incremental results rather than campaign-reported ROAS. Performance Max is frequently part of that reset, either tightened with the controls described here or paused until the account has the signal to support it.