If you’ve been working with Google’s Performance Max campaigns (PMax) for any length of time you’ve probably experienced the rug pull. Campaigns will often do well at first in, but then something will change drastically. We’ve seen campaigns crater in spectacular ways, without significant changes to the campaign, account, competition, or market demand. PMax will just suddenly drop down to half of it’s previous ROAS with no apparent cause… until you dig deeper.

If you’re in this, you’re not alone. Check out this discussion on LinkedIn where other power users are seeing very similar things.

The rest of this post is going to be recommendations on steps you can take to diagnose rug pulls. First, to level set, we have to understand that the purpose of PMax is to make better use of Google’s remnant inventory. This includes placements like: Gmail, discovery feed on Android devices/Chrome browsers, Google Display Network, low intent search terms, YouTube, and apps. In short, inventory that doesn’t work well on it’s own. Google’s been trying to sell this inventory for years and has finally found a path. PMax uses machine learning to maximize it’s sales of this inventory while trying to deliver performance to the advertiser. I leave it to you, dear reader, to decide which of those two things has priority with Google’s shareholders.

Diagnostic step 1. Are the tags set correctly?

In the Google ads interface click on the wrench icon for settings. Navigate to Audience Manager and find the selection on the left “My Data Sources”.  Look for the Details link for the Google Ads Tag.
Google Ads Tag Details

In this detail link you’ll see what parameters Google is receiving. If you’re not sending the right parameters PMax won’t be able to do retargeting correctly. Something between 10 – 30% of most PMax campaigns is retargeting.

Many platforms like Shopify, Magento, and BigCommerce are not sending the right details out of the box. You’ll need configure your Gtag event snippet to send product id (ecom_prodid has been deprecated) like the example below.

gtag('event','view_item', {
  'value': 998.55,
  'items': [
      'id': 1234,
      'google_business_vertical': 'retail'
      'id': 45678,
      'google_business_vertical': 'retail'

Diagnostic step 2. How much of your campaign spend is going to shopping?

Mike Rhodes has developed a script that he has generously shared with the Google Ads community that can show you exactly how much of your spend is going to Shopping compared to other Google channels.
PMax Campaign Spend Percent Shopping
In some cases 50% or more of the campaign can be going to other channels. Google has a problem with click fraud. It’s doing a better job at detecting fraud in search than it does in display. If you look in the PMax placements you may see that lots of apps or “made for advertising” websites are getting clicks. If this is happening you can experiment by making no asset campaigns to force PMax to focus on Shopping.

Diagnostic step 3. How much of your spend is going to irrelevant products?

For most of us we have a few products that drive the most revenue, but we often find that Google’s algorithm is focused on less important products. Dive into your product report. Sort by impressions, then sort by cost. If you see that the wrong products are being promoted you’ll need to segment them out. Don’t wait for the algorithm to figure it out. You could go broke before the algorithm understands your business or your customer.

Diagnostic step 4. Are you showing for relevant searches

Search terms aren’t available in PMax. One of the best things we have as an indicator is the Auction Insights Report.

Shopping Campaigns Auction Insights

If you’re showing against relevant direct competition, you’re probably showing for relevant searches. If your main auction insights overlaps are with Amazon and Walmart you might not be.


Segmentation – What products should be treated special? Will some products benefit from the additional reach of PMax? What about cheap frequency? If the product has done well with remarketing in the past, it will probably do well with additional assets. If not, it should be segmented into a no-assets campaign. What about margin or profitability? Think about segmenting along ROAS or profitability goals.

Hero products – You may want a few hero products in PMax and the rest in standard shopping. Hero products often benefit from additional visual assets. Placing the rest of the products in standard shopping will allow you to get the keyword data associated with the clicks and use that in your overall marketing plan.

Block Apps – PMax attempts to maximize the use of Google inventory. Apps are cheap inventory but apps are also notorious for fake or fraud clicks. Even for valid clicks most apps don’t perform well for ecommerce product campaigns. Google changes what you have to do to block apps every year. As of 11/22/22 to block all apps you’ll need to block each of the 140 categories of apps one by one.

To block all apps, navigate to the Tools and Settings wrench icon and find Content Suitability.

Google's Content Suitability

Expand Excluded Placements and expand App Categories. You’ll need to select every category and save.

App Category exclusion

Run Standard Shopping campaigns – We still like Query Filtered (keyword targeted) Google Shopping campaigns in most cases. All of the spend goes to shopping. A lot more data is available for optimization. We can use bid triggers based on margin, inventory assortment, or new products. Google is terrible at figuring out a product is new and giving it a bit of a boost until it gets traction. Standard shopping campaigns aren’t going away anytime soon. The added benefit of running a standard shopping campaign, especially early on, is you can train the data for PMax.

For both standard shopping and PMax all keyword performance data is tied to the product identifier. If we bid up a product for certain keywords, that keyword performance becomes tied to that product. It can give PMax a big jump start.

Standard shopping is the ultimate way to avoid the rug pull or deal with it when it comes.