Performance Max campaigns frustrate many Google Ads marketers for being a black box.
These AI‑driven campaigns can serve ads across Search, Shopping, Display, Discover, Gmail and YouTube, yet marketers often struggle to extract insights and scale them efficiently. However, more data and features have become available recently.
In this article I’ll go over some useful tactics to optimize and scale PMax campaigns.
2 Important Points Before We Start
Data Is King
Performance Max campaigns rely heavily on Google’s AI bidding model. For this to generate good results, it needs a lot of data.
Before you make any changes, make sure that you have enough data. That is, enough data for the bidding algorithm, and for you to make data-based decisions.
If your campaign uses a tROAS bid strategy, Google recommend at least 15 conversions in the last 30 days, but Google Ads specialists insist on at the least 50. If your campaign uses tCPA, it’s best to have at least 30 conversions in the last 30 days.
Also allow enough time and conversions data after the changes, to evaluate their impact. The more conversion data your PMax campaigns have, the better.
Testing Mentality
Approach scaling as a series of controlled experiments:
- Assign changes to one campaign at a time while keeping others as controls
- Document each hypothesis in a spreadsheet and allow sufficient time and conversions before judging results
- Starting with smaller campaigns minimizes risk; once a tactic proves effective, roll it out to larger budgets
- Be critial towards test results – consider conversion fluctuations, promotions, and conversion lags when you evalutate the winner
This disciplined testing mentality underpins all the tactics below.
1. Product Feed Optimization
For e‑commerce, the product feed is the heart of your PMax campaign. By enriching product titles and descriptions with keywords, attributes and unique selling points (USPs) you can widen the range of queries that trigger your ads.
For example, instead of a generic title like “Chocolate Bar Teddybear”, a more descriptive title – “Sugar‑Free Chocolate Bar Teddybear for Kids & Toddlers, Organic, Fair‑Trade, 4 oz” – signals quality and intent. Include synonyms, materials, sizes and benefits that shoppers search for.
Ensure that the Google product categories are as accurate as possible to help the algorithm deliver your ads to the right users.
Use custom labels to tag products by margin or seasonality, enabling PMax to allocate budget intelligently. Regularly update the feed to reflect stock changes and promotions.
2. A/B Test Audience Signals
Audience signals in Performance Max campaigns are suggested by Google primarily to target users in the top and middle of the funnel.
It’s best practice to include your exisiting customers as the seed audience signal in PMax campaigns. The larger this audience is, the better Google will target new prospects. Note, that audience signals are different from targeting. Adding your existing customers as a signals doesn’t mean Google will only show your ads to them, rather try finding similar users.
You may target affinity, in-market, custom, and remarketing audiences. However, there are very few insights in the Google Ads UI as to how these audiences perform in terms of spend and conversions.
So one way to test this would be to create 2 asset groups with the same SKUs but with different audience signals. For example, one asset group with remarketing audiences, and another one without. Then, you can allow these 2 asset groups to run for a couple of weeks and see whether one performs better than the other.
3. Raise Budgets
If your PMax campaign hits the target and is limited by buget, you can raise the budget gradually, no more than 20% raises at a time. This will prevent your campaign from re-entering the learning phase.
If your campaign is too far from its target ROAS or target CPA, increasing the budget will probably have little effect, if any.
Increasing budget will often result is decreasing efficieny (low of diminishing returns). By how much, is a question of your position on the demand curve.
4. Adapt Targets
If your campaign’s actual tROAS or tCPA is below the target, and overall account performance is acceptable, you could adapt your bidding target. Consider decreasing the target to be ca. 10% below actual performance.
E.g. if actual ROAS is 300%, decrease the target to 270%; if actual CPA is $50, increase your target to $55. If these targets are far from your current ones, consider adapting gradually.
From my experience, too low bids in PMax result in higher spend on TOFU channels, i.e. YouTube and Display, which are typically less efficient than Shopping ads.
Google Performance Planner can help estimate the impact of bid and budget changes on performance.
5. Segmentation By SKUs
In most Google Ads accounts, only a few SKUs drive most spend and revenue. Leave these top performers in the original campaign, and build new campaigns for the rest of your SKUs.
You may want to segment those by category, performance, or margins. The more data you have, the more campaigns you can build.
Tip: If you’d like to scale SKUs with a similar AOV, you can also test setting them in a separate campaign with a tCPA bid strategy. If your SKUs are priced very differently and your objective is revenue and profit, tROAS is more appropriate.
6. Focus On Best-Sellers
Similar to the previous point, but without building new campaigns.
Assuming you have a few best-sellers outperforming the rest, let them run alone. After a while, the campaign’s performance should improve against the target, allowing higher spend.
7. Test A Feed-Only PMax Campaign
One of the main problems advertisers see with Performance Max, is that one can’t control how much is spent on the different ad formats. I’ve seen campaigns where 30% of the budget is spent on video ads, which perform far worse than shopping ads.
One way to control this would be to create feed-only performance max campaigns. This means, without image and video assets.
The only way this works at the moment is by creating the new campaign in Google Merchant Centre, and then editing the targeting, bids, and asset group (SKUs) in Google Ads. If you already have a feed-only asset group in another campaign, you can also duplicate it via the Google Ads editor and adapt to the relevant products.
This is the closest thing to a standard shopping campaign. So why not run standard shopping? The only advantage I see here for PMax would be if you’re using a tCPA bid strategy.
8. Optimize Assets
I left this one to the end, as from my experience it’s hard to measure its impact on performance.
Essentially, you can review your assets performance report, and replace all assets that rank ‘poor’ with new ones.
If you have new images, videos, or titles that you believe can have more potential, you can also add them to the asset group.
Another common practice is to let 2 asset groups run in parallel with different assets (and same SKUs). This often targets different users.
After a few weeks of collecting data you can pause the worst performing asset group.
9. Add Negative Keywords
In mid 2025, Google has finally made the search terms report avilable in PMax campaigns. Additionally, it enabled adding negative keywords.
Make sure you review the search terms report every now and then and exlude irrelevant and poor performing search terms.
Use this to add guardrails and exclude irrelevant, or poor performing search terms.
Conclusion
While Google’s Performance Max campaigns offer limited insights and rely heavily on AI and automation, there are ways to scale them. Product feed optimization and adapting bids are just a few.
Make sure the campaigns collect enough data before making changes or drawing conclusions, and employ a testing mentality.
Further Reading:
- Performance Max Best Practices
- 13 Ways To Optimize Google Shopping Ads (In A Profitable Way)
- Examples of Using ChatGPT for Google Ads
- Demand Gen: Google Ads Version of Paid Social Campaigns?
- Google Ads Bidding Strategies: The Ultimate Guide (And Which One To Use)
- How To Increase Conversion Rate On Shopify
