When it comes to Performance Max best practices, many ecommerce brands are still working from advice built for an older version of PMax. However, a lot has changed in this area over the last few years, and if you’re basing your strategy on practices from even one or two years ago, you’re already behind.
Google has changed how the auction works, added campaign-level negative keywords, doubled the number of search themes to 50 per asset group, and rebuilt reporting from the ground up. The ecommerce brands seeing stronger conversions and ROAS are building out their setup to match what the platform can actually do now.
One solid piece of evidence for this can be seen in a case study published by Google. A shoe brand swapped out the default, auto-filled creative in its Performance Max campaign for real video and lifestyle imagery, and the results moved fast: conversions up 60% and revenue up 59%.
Performance Max works differently from any campaign type Google has built before it. Instead of picking keywords or placements, you hand it your product feed, creative, and a goal, and the algorithm decides who sees what and where. One campaign now does the work that used to take five separate ones, pulling from the same feed and budget instead of five disconnected setups competing for the same customer.
How well that setup performs comes down to how well it’s built. Here are the seven practices we’ve found make the biggest difference in campaign performance.
A lot ecommerce stores default to splitting campaigns by category, like “shoes” and “bags.”
Here’s the problem with that: say your “Shoes” campaign has some products at a 60% margin and others at a 10% margin. Since one campaign only gets one ROAS target, the algorithm treats every shoe the same, and it pushes budget toward whatever sells fastest, usually your already-popular products, whether or not those are the ones actually making you the most money.
Splitting too early causes a second problem. Five or six campaigns competing for only 50 total conversions a month means each one gets too little data to learn from and underperforms. Wait until you’re past 100 conversions a month before splitting anything.
Once you hit that point, here’s a structure that actually reflects how your catalog makes money instead of how it’s categorized:
Your proven, ongoing product range
ROAS target reflecting blended margin
Enough performance history to support a stable ROAS target.
The 10 to 25 SKUs that typically drive 70 to 80% of revenue
Dedicated budget with a tighter ROAS target
Your biggest revenue and profit opportunity deserves its own campaign.
Product launches and seasonal items without performance history
Maximize Conversion Value with no ROAS target yet
There is not enough conversion data to set a realistic ROAS target.
Performance Max doesn’t let you target keywords directly. Your product feed does that job instead, and many ecommerce advertisers coming from Search underestimate just how much weight it carries.
The algorithm reads your feed data to match your products to real search queries. If your titles are vague or your descriptions are thin, your products simply won’t show up for searches you should already be winning. No amount of clever campaign structure fixes that.
Here’s what to check before you touch anything else:
Thin asset groups give the algorithm thin material to work with, and what you put in directly shapes what comes out. This shows up most clearly with video. If you don’t add one, Google auto-generates it from your other assets or your product feed, and that default creates a real gap.
A study covering more than 24,000 Performance Max campaigns found that accounts running real video assets held up better across different levels of buyer intent, and multiple agencies running Performance Max at scale report real video outperforming the auto-generated version by 25 to 40%.
That same principle is really what asset groups and creative best practices come down to across the board: give the algorithm real material, not the bare minimum. Here’s what a full asset group should include:
Audience signals don’t work the way many ecommerce advertisers assume. The algorithm treats them as a starting point, and it will still show ads to people outside your signals if it predicts they’re likely to convert. Google splits audience signals into three real categories inside Performance Max, and they carry a different connection to your actual customers depending on the type.
Here’s how to prioritize what goes into each:
1. Customer Match
A list of your actual customers built from their email addresses. Your strongest audience signal because it's based on real purchases rather than predictions.
- Segment customers by value, such as high LTV or high AOV buyers.
- Use Google's New Customer Acquisition setting with high-value customers to prioritize your most valuable audience.
- Keep your Customer Match lists up to date with regular uploads.
2. Website Visitors & GA4 Audiences
People who have visited your site but haven't purchased yet. This is still first-party data, but one step removed from an actual sale.
- Create separate audiences for engaged visitors and cart abandoners.
- Build these audiences as early as possible since GA4 only starts collecting members after an audience is created.
- Prioritize recent visitors over older traffic
3. In-Market Audiences
Groups Google predicts are actively shopping in your category based on broader browsing behavior rather than your own customer data.
- Add these only after your first-party audiences are in place.
- Use them to expand your reach rather than as your primary audience signal.
- Treat them as a supplement, not the foundation of your targeting strategy.
One mistake worth watching for: tightening your ROAS target and tightening your audience signal at the same time. Doing both at once restricts the algorithm on two fronts simultaneously, and it’s a common reason campaigns plateau right after a setup change instead of improving. Give one lever room to work before adjusting the other.
Setting a ROAS target the moment you launch a campaign feels responsible. In practice, it’s one of the more common ways ecommerce brands sabotage a new Performance Max campaign before it gets a fair chance to prove itself.
Here’s the actual problem with setting a target too early: a ROAS target tells the algorithm to chase only conversions that meet a specific return, but on day one, the algorithm has no real data yet on which auctions, audiences, or placements actually deliver that return for your account. Forced to hit a number it can’t yet predict, it ends up bidding too cautiously and missing conversions it would have otherwise found.
Sequencing your bidding strategy into two stages beats picking a single approach and locking it in from day one. Here’s how:
A reasonable bar for making that switch is 30 to 50 conversions in the past month, or at least 3 full conversion cycles, whichever takes longer. Set an aggressive ROAS target before you hit that volume, and total conversion volume can drop dramatically; we’ve seen drops as steep as 50% when a target gets forced on too early.
Once your target is live, treat it with the same patience. Google recommends moving it in increments no larger than 20% and no more often than every two weeks. Anything faster or bigger just keeps resetting what the algorithm has already learned, which means you end up paying for a second learning phase you didn’t need to trigger in the first place.
Many e-commerce brands think that doubling the budget after one good week is a strategic move. However, it’s actually one of the fastest ways to undo a campaign that was finally starting to perform.
Increasing the budget by a large percentage in one move tells the algorithm to compete in a much bigger set of auctions overnight, many of which it hasn’t learned how to bid on efficiently yet. Doubling a budget in one jump has been shown to raise CPA by 25-50% and drop ROAS for one to two weeks before it recovers.
Here’s how to scale without triggering that:
Performance Max can quietly take credit for branded search conversions your dedicated Search campaign should be getting anyway.
Search is supposed to win when a search matches an exact-match keyword in your Search campaign. In practice, that doesn’t always hold. An analysis of 503 real Google Ads accounts found keyword overlap between Search and PMax in 91.45% of them, affecting more than half of all Search campaigns studied.
Here’s what’s usually behind that overlap:
Brand exclusions fix this by stopping PMax from bidding on your branded terms at all, leaving that traffic for Search to handle instead.
Situation | Recommendation |
Branded Search performs well and you want cleaner reporting | Apply brand exclusions to your Performance Max campaign. |
No dedicated branded Search campaign exists | Create a branded Search campaign first, then apply brand exclusions to Performance Max. |
Unsure how much branded traffic Performance Max is capturing | Review the campaign’s Search Terms report before deciding whether to apply brand exclusions. |
Implementing the Performance Max best practices above can genuinely improve your campaign results. But in the process of getting the big picture right, it’s easy to make smaller mistakes that look harmless at first, and quietly hold your results back from where they should be.
Here are some of the most common mistakes ecommerce brands make with Performance Max, and how to fix each one.
A new Performance Max campaign needs a specific sequence in its first month. Launching with a ROAS target, full creative, and tight audience signals on day one usually backfires, since the algorithm needs real conversion data before those settings can actually help rather than restrict it.
Here’s a practical, week-by-week roadmap for the first 30 days:
Performance Max hands over more decision-making to Google’s algorithm than any other campaign type, which means it depends more heavily on the quality of what you feed it than a campaign where you’re manually choosing keywords and placements
Before launching or before troubleshooting a campaign that’s already live and underperforming, here’s what to confirm first:
If more than one of these isn’t in place yet, fix them before launching or scaling further. No amount of campaign structure or creative quality makes up for the algorithm working with incomplete or inaccurate data.
Performance Max has changed enough since its introduction in 2021 that much of the advice still circulating online no longer reflects how the platform actually works. Here’s what’s changed, and what it means for how you should be running campaigns now:
Performance Max only performs as well as what you feed it. A campaign structured around margin, a feed that reflects how people search, real creative instead of auto-generated defaults, and a bidding sequence that gives the algorithm room to learn: these are what turn Performance Max into a system that keeps compounding.
The hard part shows up once you start scaling, since what works at $3,000 a day gets harder to manage at $30,000 a day. More campaigns and more asset groups mean more decisions about budget and timing happening at once. Without a systematic process behind how Performance Max and the rest of your Google Ads account are structured, growth turns into inconsistency instead.
Want a more systematic way to manage Performance Max as it scales? Book a free audit with Vysta. We manage Performance Max for ecommerce brands generating seven to nine figures in revenue, and we build the structure that keeps performance compounding as spend grows.
The best Performance Max practices for ecommerce start with structuring campaigns around margin, and treating your product feed as your real targeting strategy. Once that foundation is in place, add real video instead of relying on Google’s auto-generated version, and feed the algorithm your customer list before anything else. From there, sequence your bidding so no ROAS target gets set before the algorithm has real data, and scale budget in small, steady increments to keep results compounding.
Yes, Performance Max is good for ecommerce brands because it puts your product feed to work across every Google channel from a single campaign, matching your products to real search queries automatically. It performs best once you have reliable conversion tracking, a clean feed, and enough sales volume for the algorithm to learn from.
A Performance Max campaign typically needs 4 to 6 weeks just for the initial learning phase, with full optimization and stable performance taking 60 to 90 days overall. During that stretch, Google’s AI is gathering data and calibrating bids across every network. Reaching 30 to 50 conversions a month is what actually gets a campaign to that stable point faster.
Google recommends at least 30 conversions in the past 30 days before launching Performance Max, since that gives the algorithm enough historical data to optimize bidding and targeting. Accounts still in the single digits are better off building a foundation with Standard Shopping or Search first. Beyond volume, conversion tracking needs to be firing accurately, since bad data undermines even a healthy conversion count.
Yes, you should use Standard Shopping and Performance Max together. Run Performance Max on your bestsellers to scale volume across every Google network, and let Standard Shopping handle low-margin SKUs, slow-moving inventory, and new products you still want to test manually. That split only works if the two campaigns stay out of each other’s way, since Performance Max takes priority the moment the same product shows up in both.
Start with 2 to 4 asset groups per Performance Max campaign, since Google allocates budget at the campaign level and too many groups early on spread both data and spend too thin. You can add more as performance data builds, up to 100 per campaign, once you’ve identified real segments by category, audience, or promotion that genuinely warrant separate creative. Most ecommerce accounts never need more than a handful.
The bidding strategy you should use for Performance Max is to run Maximize Conversion Value paired with a Target ROAS, since it directs the algorithm toward your highest-revenue conversions instead of treating every sale as equal. Start every new campaign with no target set, letting the algorithm collect baseline data first. Once you’ve hit at least 30 conversions in the past 30 days, add a Target ROAS based on your actual account performance.
You should optimize a Performance Max campaign on two different schedules: check performance weekly to catch anomalies in spend or search terms, but only make structural changes, budget, targets, or assets, every 3 to 4 weeks. Changing things too often resets the algorithm’s learning phase and causes volatile results. When you do adjust a target, a 10 to 20% move at a time tends to work best.
The most common causes of a Performance Max campaign not performing are insufficient conversion data, weak or minimal creative assets, and a ROAS or CPA target set too aggressively for the algorithm to work with. Below 30 conversions a month, the algorithm simply doesn’t have enough signal to optimize effectively. Final URL Expansion can also quietly waste budget by sending traffic to unoptimized pages if it isn’t reviewed.
Performance Max can’t fully replace Standard Shopping campaigns for every ecommerce account, since the two serve different purposes. Standard Shopping gives you exact search term data and product-level bid control, while Performance Max trades that visibility for reach across Search, Shopping, YouTube, Display, Gmail, Discover, and Maps. Many accounts run both together, using Performance Max for volume and Standard Shopping to protect margins on specific products.
Book a call with Nate Schneider to explore how Google and YouTube ads can drive scalable, measurable growth.
Blog SEO and Google Ads: Why the Fastest-Growing Ecommerce Brands Invest in Both For many ecommerce brands, acquiring new customers through search comes down to two channels: SEO and Google Ads. But with how competitive the space has gotten, choosing one over the other is no longer the strongest move. Winning in both is what […]
