Last updated May 2026.
You launched a Performance Max campaign for your business, the budget started spending, and now the question is whether it is actually doing what it was supposed to do.
This is the most common Google Ads conversation we have with new clients, and the answer almost always comes down to optimization. The 2024 version of Performance Max was a black box that mostly worked or mostly did not. The 2026 version of Performance Max gives advertisers a meaningful set of levers, including new-customer acquisition bidding, brand exclusions, search themes, account-level negative keywords, and channel-level reporting that finally tells you where your conversions are actually coming from. Used correctly, those levers can take a campaign from underperforming to scaling. Ignored, they leave most accounts running on the algorithm’s defaults, which are designed to maximize Google’s revenue, not yours.
This guide is a 2026 refresh of how we optimize Performance Max campaigns for our clients. We walk through what Performance Max is now, the optimization strategies that move the needle in 2026, and the configuration mistakes we routinely fix on accounts inherited from previous agencies.
If your PMax campaign is running but the numbers do not match what you were promised, an audit usually finds the answer in under an hour.
What is Performance Max in 2026?
Performance Max is a goal-based Google Ads campaign type that uses machine learning to serve a single campaign across all of Google’s inventory. One PMax campaign can show on Search, the Display Network, YouTube, Discover, Gmail, and Google Maps simultaneously, with Google’s algorithm dynamically choosing where each impression goes based on conversion likelihood.
What separates a 2026 PMax campaign from a 2024 one is the amount of control advertisers now have over what the algorithm does:
Cross-channel reach. One campaign covers Search, Display, YouTube, Discover, Gmail, and Maps. As of 2025, Google reports performance breakdowns by channel inside the campaign view, which means you can finally see which surfaces are producing your conversions instead of guessing.
Smart bidding by default. PMax requires a smart-bidding strategy. The choice is between Maximize Conversions (with optional Target CPA) and Maximize Conversion Value (with optional Target ROAS). The algorithm uses real-time signals to decide what to bid for each individual auction.
Asset groups instead of ad groups. Each PMax campaign contains one or more asset groups, which are bundles of headlines, descriptions, images, videos, and logos that Google mixes and matches. Asset group reporting got significantly more granular in 2024 and 2025, allowing advertisers to see how individual assets perform.
Audience signals and search themes. Both are suggestions, not targeting. Audience signals tell Google which customer segments resemble your buyers. Search themes, introduced in 2023, let you suggest specific search queries you want PMax to compete for. The algorithm uses these as starting points and then expands.
Brand controls and customer-acquisition controls. The two newest meaningful additions. Brand exclusions let you stop PMax from cannibalizing your branded search traffic. New-customer acquisition bidding lets you tell Google to weight new customers as worth more than existing ones, with two flavors covered later in this guide.
Used together, those features make PMax much closer to a real optimization surface and much less of a black box than it was even 18 months ago.
How to evaluate whether your PMax campaign is performing
Before you start changing settings, decide whether the campaign actually needs optimization. We start every PMax audit by checking three things in order.
First, are conversions tracking accurately? More than half of the underperforming PMax accounts we audit have a tracking problem rather than a campaign problem. Open the conversion action in Tools and Settings, confirm that primary conversions match revenue-producing actions like booked calls, completed purchases, or qualified leads, and verify that lower-bar actions like clicks-to-the-phone-number or add-to-cart are not double-counting. PMax bids on whatever you tell it counts. If the conversion definition is wrong, every other optimization is built on sand.
Second, what does channel-level reporting say? Open the campaign, switch to the Insights or Channel Performance view, and look at where impressions, clicks, and conversions are coming from. A campaign serving 80 percent of impressions on Display with very few conversions is a different problem than a campaign serving 80 percent of impressions on Search with high CPCs.
Third, what is the new-customer split? If you have new-customer acquisition bidding configured (covered below), the campaign view shows new-vs-existing customer conversions side by side. If almost all conversions are existing customers, the campaign is operating as a remarketing engine, not an acquisition engine, and the fix is bidding-related, not creative-related.
Once you have those three answers, the optimization strategies below become specific levers you pull on purpose, instead of changes you make hoping something will work.
Optimization strategies for Performance Max in 2026
The order below is the order we work through PMax accounts, from the foundational structural decisions down to the day-to-day adjustments.
Account and campaign structure
The single most common structural problem we see is one giant PMax campaign trying to cover every product, every audience, and every region at the same time. Because PMax shares budget across the entire campaign, that structure forces the algorithm to spend wherever it finds the cheapest conversion, which usually means the highest-margin or most price-sensitive segment of your business gets starved.
The fix is to split campaigns by economic logic, not by product taxonomy. We typically run separate PMax campaigns for: branded vs. non-branded, top-margin product lines vs. lower-margin lines, prospecting vs. retention, and major geographies. The reason to split is so each campaign has a homogeneous goal, which lets bid strategy and budget pacing actually do their job.
For multi-location local businesses, we also override at the campaign level for location assets, so each campaign features the right office. We covered this in our guide to call and location assets, which pairs naturally with PMax for any local business.
Budget pacing
Performance Max has unlimited inventory across Display, YouTube, Gmail, and Discover. Unlike standard Search campaigns where impression share is capped at 100%, PMax can almost always spend more than you give it. That makes budget the single most powerful lever in the campaign and also the easiest one to misuse.
Increase budget in increments of 10 to 20 percent every two to three days during scaling phases. Larger jumps trigger Google’s relearning period, where the algorithm re-evaluates how to spend the new budget and conversion volume usually drops for one to two weeks. We have watched accounts double their budget overnight, lose half their conversion volume for ten days, and then watch the founder panic and pull spending back down before the algorithm could recalibrate. Slow scaling is faster than fast scaling.
For decreases, the rule is the same. Pull spend down by 10 to 20 percent at a time. Yanking a budget by 50 percent in a day signals the algorithm that the campaign goal has materially changed, and the relearning period kicks in just as harshly as it does for increases.
Bid strategy and targets
PMax requires a smart-bidding strategy. The choice is between Maximize Conversions and Maximize Conversion Value, with or without a target.
For lead-gen accounts where every conversion is roughly equal in value, Maximize Conversions with a Target CPA works well. For e-commerce or any account where customers spend wildly different amounts, Maximize Conversion Value with a Target ROAS is almost always the right answer because Google can factor differential conversion values into bidding decisions instead of treating a $20 sale and a $200 sale as equal.
Even for lead-gen, we usually move clients to Maximize Conversion Value once we can assign meaningful values to different conversion types. A booked-call conversion is worth more than a price-guide download. A demo request is worth more than a newsletter signup. Telling Google those relative values gives it more signal to work with.
For target CPA or ROAS values, do not set unrealistic targets. If your historical campaigns have produced leads at $80, setting a target CPA of $30 will throttle the algorithm so aggressively that it stops bidding at all. Set targets within 20 percent of your trailing 30-day actuals, and adjust no more than once every 14 days. Google calculates target performance on a 30-day window, and faster adjustments just keep the algorithm in a constant relearning state.
New-customer acquisition bidding
The single highest-leverage 2026 setting in PMax. Inside the bidding section there is an option called Customer Acquisition. When you turn it on, you choose either “Bid higher for new customers,” which keeps existing-customer conversions but values new-customer conversions more, or “Bid only for new customers,” which effectively turns off bidding on existing-customer auctions.
For nearly every business, “Bid higher” is the right call and “Bid only” is a trap that costs accounts revenue. The mechanics, the calculation for setting the incremental customer value, and the rare exceptions where “Bid only” makes sense are all in our dedicated guide to Performance Max new-customer acquisition bidding. If your campaign feels like it is just hammering existing customers, this is the setting to fix first.
Brand exclusions
Performance Max defaults to chasing the cheapest conversions, which often means it bids on your own brand keywords because branded searchers convert at very high rates. The result: PMax cannibalizes the conversions that would have come through your brand search campaign anyway, while charging you for clicks you would have gotten organically or at a much lower CPC.
The fix is brand exclusions. Inside the Performance Max campaign settings, navigate to “Brand exclusions” and add your branded terms (your business name, common misspellings, branded product names). PMax will stop bidding on auctions where the search query matches those terms, which both protects your brand search campaign’s data and prevents budget from being wasted on traffic you would have captured anyway.
For accounts running both a branded Search campaign and a PMax campaign, brand exclusions on the PMax side are non-negotiable. We add them on every account we manage during onboarding.
Search themes and audience signals
Audience signals and search themes are both suggestions to the PMax algorithm, not hard targeting. The distinction matters: audience signals tell Google which customer segments resemble your existing buyers, while search themes (introduced in 2023) let you tell Google which search queries you want to compete for.
For audience signals, the highest-quality input is a customer-match list of your actual customers, ideally segmented by lifetime value. Lower-quality inputs are interest-based or demographic segments that you assume describe your buyers. We always recommend starting with first-party data and only layering in lookalike segments after the campaign has had time to learn from the customer match list.
For search themes, list 5 to 15 specific queries you want PMax to bid on. These should be the high-intent terms a customer would actually type, not generic category labels. “best dental implants near me” is a useful search theme. “dentistry” is too broad to help. Search themes do not lock the campaign to those queries, but they meaningfully shape where Google starts looking.
Asset groups and creative testing
Asset groups are the creative core of any PMax campaign, and the asset group reporting Google rolled out in 2024 and 2025 is the single biggest improvement to the platform since launch. Open any asset group, sort by asset type, and you can see how individual headlines, descriptions, images, and videos perform on conversion value over cost.
The optimization process is straightforward once you have the data. For each asset type, identify the top performers and the bottom performers. Pause or remove the bottom performers. Build new variants modeled on the top performers, with the same color palette, framing, or messaging. Repeat the cycle every 30 to 60 days.
The single biggest creative-side win we see in PMax is adding video assets to asset groups that did not have them. Google heavily rewards complete asset group coverage in its ad-strength scoring, and the YouTube-eligible inventory unlocks meaningful incremental impressions. Even short, simple vertical videos cut from your existing photo assets can move the needle. Google’s recommendations to add horizontal and square videos are usually correct and we accept them by default.
Negative keywords at the campaign and account level
For most of PMax’s history, the only way to add negative keywords was through Google rep tickets. That changed: campaign-level and account-level negative keyword controls are now available directly inside the Google Ads UI for most advertisers as of 2024-2025.
Our approach: account-level negatives for blanket exclusions you never want to bid on (competitor brand names, irrelevant industries, employment-related terms like “jobs” or “salary”), and campaign-level negatives for terms specific to a single campaign’s mission. We also use the search terms report inside PMax Insights to find the wasted-spend queries to add as negatives.
For e-commerce specifically, common negative themes include free, cheap, used, DIY, repair, and any city or region you do not service. For services, common themes include jobs, careers, salary, free, and DIY equivalents. We covered the broader logic of how negative keywords increase conversions in a separate guide that pairs well with PMax tuning.
Ad scheduling and bid adjustments
PMax has limited bid-adjustment options compared to standard Search campaigns. The one that does work and matters most is ad scheduling. If your business operates only during certain hours, set the campaign schedule so PMax pulls back during off-hours instead of paying for clicks that go nowhere.
For local-services businesses with phones that are not answered after 6 p.m., scheduling alone routinely cuts wasted spend by 8 to 15 percent. For e-commerce, scheduling matters less because customers buy at all hours, but for retailers with margin-sensitive promotions, scheduling around the highest-conversion hours can lift overall ROAS without changing bid targets.
Product feed and Google Merchant Center
For e-commerce PMax campaigns, the product feed is roughly 60 percent of the campaign’s actual performance lever. A clean, complete, accurate Merchant Center feed unlocks Shopping placements; a feed full of disapprovals or limited-status products quietly suppresses traffic regardless of bid strategy.
Check Merchant Center daily for product status issues and resolve them within 24 hours of disapproval, because the algorithm reduces visibility on accounts with high disapproval rates even after issues are fixed. Use custom labels to segment products by margin, season, or promotional status, so PMax can be told which products are higher-priority through asset group structuring or separate campaigns.
The most common feed issues we fix on inherited accounts are: missing GTINs, mismatched product titles vs. landing-page titles (which Google flags as policy violations), missing high-resolution images, missing structured data on the source pages, and category misclassification. None of these are dramatic individually, but they compound, and a clean feed almost always lifts campaign performance more than any bid-strategy change.
Most accounts have three or four of these settings configured wrong. We can show you exactly which ones in a 30-minute call.
What we have seen with our clients
The accounts we inherit fall into three predictable categories.
The first is the underconfigured account, where PMax was launched with default settings, smart bidding pointed at the wrong conversion action, no brand exclusions, no search themes, no asset group reporting reviewed, and the customer-acquisition setting untouched. These accounts usually look like they are spending fine but converting poorly, because the algorithm is being asked to optimize the wrong thing. The fix is sequential: tracking, then bidding, then brand exclusions, then customer acquisition, then asset groups. We typically see 30 to 50 percent improvement in cost per conversion within 60 days.
The second is the overconfigured account, where the previous agency added so many search themes, audience signals, brand exclusions, and negative keywords that the algorithm has almost no inventory left to optimize against. These accounts feel safe but cannot scale. The fix is paring back to a smaller, cleaner configuration and letting the algorithm have room to learn. We have documented this exact pattern in our e-commerce Google Ads case study, where simplifying a heavily-restricted PMax campaign produced a 3x return on ad spend.
The third is the local-services account, where PMax is running but the call asset, location asset, and Google Business Profile connection are misconfigured or missing entirely. For these clients, the conversion-bottleneck almost always lives outside the bidding strategy. Once the assets are configured correctly, the same campaign produces 25 to 60 percent more phone-call conversions. The pattern is the same one we documented in our Sage Dental implant case study and our 21 new patient appointments case study, both driven by PMax campaigns paired with proper asset configuration.
Frequently asked questions
How long does it take for Performance Max to start performing well?
In our experience, the algorithm needs roughly 14 days of consistent budget and conversion data before it stabilizes, and 30 to 45 days before it produces results worth evaluating. Accounts that change targets or budgets weekly never finish a learning cycle and tend to underperform indefinitely. Our standard rule for clients is no major changes for the first 30 days after launch.
Can I run Performance Max alongside a regular Search campaign?
Yes, and we usually do. The two campaign types serve different roles. The Search campaign protects your branded queries and high-intent non-branded keywords with explicit control. PMax fills in the rest of the funnel across YouTube, Display, Discover, Gmail, and Maps. The key configuration is brand exclusions on the PMax side so the two campaigns do not compete with each other for branded clicks.
Why is my PMax campaign spending most of its budget on Display?
Almost always because Display is where the cheapest impressions are, and the algorithm is doing exactly what you told it to do: maximize conversions or conversion value at the lowest cost. If those Display impressions are not converting at the same rate as Search, the fix is usually a stricter Target CPA or Target ROAS, which forces the algorithm to be more selective about cheap-but-low-converting placements. The 2025 channel-level reporting makes this analysis straightforward.
How many asset groups should a Performance Max campaign have?
For most accounts, three to five asset groups per campaign, organized by audience or product theme, gives the algorithm enough creative variety without splintering data. Single-asset-group campaigns work for very small inventories. Campaigns with 10+ asset groups usually do not have enough conversion volume per group to optimize meaningfully and end up underperforming the consolidated version.
Should I use Performance Max if I am running a tight budget?
PMax has a practical floor of around 30 conversions per month for the algorithm to use its bidding signals meaningfully. Below that, the campaign tends to thrash because the data is too sparse to optimize against. For accounts spending under $1,000 per month, we usually recommend starting with a tightly-targeted Search campaign first and only graduating to PMax once the conversion volume justifies it.
How do I know if my PMax campaign needs to be rebuilt vs. optimized?
If the campaign has been running for more than 60 days, has burned through more than $5,000, and is still hitting your target CPA or ROAS less than half the time, the bidding signals it has accumulated are usually counterproductive and a clean rebuild outperforms continued optimization. Below that threshold, optimization almost always wins. The judgment call is whether the campaign’s history is helping or hurting future performance, and the asset-level and channel-level reporting introduced in 2024-2025 make that question much easier to answer than it used to be.
Where to go from here
Performance Max in 2026 is a meaningfully different product than it was in 2024. The optimization levers that exist now (new-customer acquisition bidding, brand exclusions, search themes, account-level negative keywords, asset-level reporting, channel-level reporting) make it possible to run PMax as a deliberately-configured campaign instead of a black-box bet on Google’s algorithm.
The accounts that win are the ones that work through these settings sequentially, give each change time to settle, and read the data instead of guessing. The accounts that lose are the ones that either changed nothing since launch or changed everything every week.
This is the single highest-leverage area we work on with our clients, and it is one of the foundational components of our broader Google Ads management service. If your PMax campaign has been running on autopilot for more than 90 days, that is almost always where the next round of growth is hiding.
Tell us about your campaigns and we will show you exactly what optimizing is worth in your specific account.
