What’s Causing Your Performance Max Campaigns to Underperform
Performance Max underperforming? Learn which data layer and tagging changes feed Google Ads bad conversion signals, and how to catch them early.
Advertising platforms keep moving decisions away from the media buyer, instead preferring to let AI make decisions autonomously. Google is no exception: look at Performance Max (pMax) and AI Max. I often hear these campaign types at the receiving end of frustration from marketers (spend 5 minutes on Reddit if you haven’t heard this).
Common Marketer Frustration Points with pMax
Marketers often cite the black box nature of the algorithms and an influx of bad traffic/non-qualified leads wasting their spend. Diving deeper, though, two common threads emerge: making some type of significant change, or not having a handle on their data.
- Frustration: “It’s a black box”
- Potential cause: It’s true we don’t know the exact reasons why decisions are made. However, changes in conversion goals, secondary actions, or optimization targets heavily influence how the campaigns are making decisions.
- Frustration: “The traffic is bad”
- Potential cause: If early, the campaign could still be learning. But as the campaign progresses, if traffic isn’t converting it could point to underlying issues with the signals being sent back to the campaign. For example, is it getting poor conversion signals (or no conversion signals)?
- Frustration: “The leads are unqualified” or “The customers don’t stay”
- Potential cause: Assuming you’re out of the learning period, this points to the data likely not being rich enough. All leads/purchases will be treated as equal unless more nuance is provided such as lead quality, purchase size, or predicted value.
- Frustration: “We switched to pMax/AI Max and our campaigns tanked”
- Potential cause: Even more so than traditional campaign types, underlying issues with conversion data have a major impact. You may have really dialed in creative or audience targeting, and now that these decisions are being made more autonomously poor signal quality is having more of an impact.
Making Sure Your Conversion Signals Are Healthy
Many of the issues come back to data quality, specifically conversion signal quality. Two factors matter here: (1) whether conversions are being recorded accurately, and (2) whether there is enough context within each conversion to train the models effectively.
When a week or more of conversion data is affected, Google notes performance can stay unstable for one to two conversion cycles. Depending on your industry, that can mean days or several weeks. So keeping on top of signal health is key, even if your campaigns are performing well right now.
How to Test if Conversions Are Being Captured Accurately
Knowing if your tags are firing accurately is straightforward, though challenging to stay on top of. This is especially true if you are managing campaigns across many different websites, brands, or regions.
If you’re operating on one or two sites, the basic troubleshooting/testing guides each tag manager provides are a good starting point (such as this for Google Tag Manager).
However, if you manage:
- Large budgets that weeks of underperformance would hurt
- Many websites
- Many regions
Then it’s worth exploring an automated monitoring solution like InfoTrust Suite to catch outages or anomalies in your conversion data.
How Broken Data Layers Feed pMax Poor Signals
The data embedded within the conversion is much more challenging to get a handle on. InfoTrust Suite enables building rules to monitor it, but you need to know what to validate.
Each event carries fields that get matched against specific requirements within Google Ads. When that field changes, goes missing, gets misformatted, or breaks in some way, the conversion will still fire so things appear to be operating normally on the surface.
Before
- Data Layer field
- value
- TMS variable
- Populated
- Conversion
- Fires
- Smart Bidding
- Healthy signal
After a case change
- Data Layer field
- Value
- TMS variable
- Empty
- Conversion
- Still fires
- Smart Bidding
- Optimizing on bad signals
Key Data Layer Fields to Monitor
It’s important to note that within each of these, the value may still be “accepted” by Google, but not be feeding the right data. For example, you might be sending a $0 value, which would still be “accepted,” even if it’s not accurate.
value,currencyTriggering event: Purchase/lead eventUsage: Smart Bidding (tROAS, Max Conversion Value)transaction_idTriggering event: Purchase/Order/Form SubmitUsage: Deduplicationnew_customerTriggering event: PurchaseUsage: New Customer Acquisition Goalemail,phoneTriggering event: Purchase/Form SubmitUsage: Enhanced ConversionsEvent NameTriggering event: Any conversionUsage: Conversion Goalspage_view,view_item,form_startTriggering event: EngagementUsage: Audience Signals
Typical Reasons Issues Occur
Governing marketing data is unique, especially for large organizations and sophisticated marketers, as it doesn’t have one true owner. Performance marketers have requirements to feed their campaigns, analysts need other data points to evaluate journeys, and UX designers need different data entirely. Meanwhile, engineering maintains the data layer, privacy teams determine usage, and analytics teams deploy the tags.
That’s a lot to go wrong.
I thought it would be helpful to bring forward some of the most common issues we see across our customers.
Select a row for details, or a column to filter.
Field name changes case (value becomes Value)Smart Bidding: affectedDeduplication: not affectedEnhanced Conversions: not affectedNew Customer Goals: not affectedAudience Signals: not affected
Causes the TMS variable to be empty.
User journey changes, event not pushedSmart Bidding: affectedDeduplication: not affectedEnhanced Conversions: not affectedNew Customer Goals: not affectedAudience Signals: not affected
The event isn’t pushed to the Data Layer, or the TMS trigger isn’t updated (e.g. the add to cart flow changes).
Content Security Policy headers block collection requestsSmart Bidding: affectedDeduplication: not affectedEnhanced Conversions: affectedNew Customer Goals: not affectedAudience Signals: affected
Requests from non-approved domains (e.g. google.com/ccm/collect) are blocked, so completed orders aren’t received by ad platforms.
Cookie prefix field cleared on the remarketing tagSmart Bidding: affectedDeduplication: not affectedEnhanced Conversions: not affectedNew Customer Goals: not affectedAudience Signals: affected
The Google Ads remarketing tag can’t read the Conversion Linker’s click ID cookies.
Template changes drop key variables from the eventSmart Bidding: affectedDeduplication: not affectedEnhanced Conversions: not affectedNew Customer Goals: not affectedAudience Signals: affected
Website or product page templates change (e.g. variant details dropped from the add_to_cart event).
Client and server-side send differing transaction_id valuesSmart Bidding: affectedDeduplication: affectedEnhanced Conversions: not affectedNew Customer Goals: not affectedAudience Signals: not affected
Missing transaction_id values do this too. Either way, conversions are duplicated.
Solutions for Ongoing Data Layer & Signal Health Monitoring
You probably already know that some of these issues will be surfaced in platform, but others are much harder to detect. You’ll find out fairly quickly if your conversions stopped altogether, but it may take deep investigation to uncover that you’ve stopped sending the products a user left in their cart in your abandoned cart event.
Check the Data Quality Before Abandoning Your Strategy
Campaign types that remove much of the human decision making and hand it to AI are beholden to the data feeding them, for better or for worse. This can be a major advantage if your data quality is buttoned-up, allowing you to outperform your competitors by simply getting the fundamentals right. Ignore those fundamentals, though, and you could be burning through your budget with little insight into why your campaigns are underperforming.
How we help
The InfoTrust Suite monitors your conversion events and data layer fields across every site, brand and region. When a field goes missing, a value changes format, or a conversion stops firing, the Suite surfaces it for your team, so you can fix it before weeks of spend run on bad signals.
Explore InfoTrust SuiteThe Suite surfaces it when
- A field goes missing
- A value changes format
- A conversion stops firing
Is Bad Conversion Data Holding Back Performance Max?
The InfoTrust Suite checks your conversion events and data layer across every site, so issues surface before weeks of spend run on bad signals.
Book a Demo
