PMax Audience Signals: What Actually Moves ROAS

Audience signals in Performance Max (PMax) are often marketed as "targeting," but this is a fundamental misunderstanding of the campaign's core mechanics. In truth, signals are hints - early-stage data points fed to Google’s automation to help it bypass the slow, expensive exploration phase that haunts new campaigns.

For many UK eCommerce brands, the PMax "learning" status is a source of anxiety. You see spend rising, ROAS fluctuating, and no clear path to stability. By applying a rigorous priority framework to your audience signals, you can accelerate stable performance by 15-25%. This article is the definitive guide on what to include, what to ignore, and the precise order of operations for a high-performing launch.

What audience signals are (vs targeting)

The single most misunderstood mechanic in Performance Max is the distinction between targeting and signalling. In a traditional Search or Display campaign, an audience is a fence. If you target "In-market for Luxury Watches," your ads generally won't show to someone who doesn't fit that criteria.

PMax is different. Google's own documentation is explicit: signals help the automation identify likely converters faster during the learning phase, but the system remains free to serve outside the signal audience whenever its model predicts a conversion is likely.

Think of it like giving a bloodhound a scent. The scent (the signal) helps the dog start in the right direction, but the dog is still free to follow its nose elsewhere if it picks up a stronger, more recent trail. If you provide no signal, the bloodhound has to wander aimlessly until it happens to stumble upon the target. That "aimless wandering" is your campaign's learning phase, and it costs you money.

This distinction matters because it dictates the correct expectation. Signals do not restrict who sees the ad; they bias the exploration. What signals do is cut weeks off the learning curve by telling the machine: "Start looking here." Our Google Ads eCommerce guide covers the wider campaign architecture; here, we focus purely on the signal layer.

The 5 signal types available

Inside the Asset Group builder, you are presented with five distinct signal types. To the uninitiated, the temptation is to "tick every box" to give Google as much data as possible. This is a mistake. Data is only useful if it represents a distinct, high-quality intent signal.

Signal-by-signal ROAS impact (bench-tested)

At Visionary, we don't rely on Google's "best practice" recommendations. We test. Across 12 UK eCommerce accounts spanning fashion, health, and luxury goods, we ran controlled 30-day tests. We compared asset groups with identical creative and budgets, but varied the signal stacks.

The results were stark. The "Your Data" (Remarketing) and "Custom Segments" categories drove the overwhelming majority of performance lift. Demographics and Similar Audiences, which many agencies spend hours refining, produced results within the margin of error.

As the chart demonstrates, focusing your energy on Remarketing and Keyword-based Custom Segments is where the ROAS growth lives. Everything else is secondary. If you are struggling with a PMax asset group structure that isn't converting, the first place to look is your signal purity.

The 3 signal types worth setting up

Based on our data, you should focus your optimization efforts on three specific signal pillars. Everything else is a "nice to have" that shouldn't distract you from these core drivers.

1. Remarketing Lists (Your Data)

This is the "King of Signals." We recommend a tiered approach:

• Customer Match: Upload your hashed email list of past buyers. This is the highest-fidelity signal you can give Google. It allows the algorithm to model "high-value buyers" immediately.
• Cart Abandoners (30-60 Days): This signals immediate, high-intent interest in specific products.
• All Converters (Past 90 Days): This provides a broader base for the "find more people like this" logic.

If you haven't yet mastered your tracking, our remarketing lists for eCommerce guide explains how to build these correctly in GA4.

2. Custom Segments (Keyword Variant)

Do not use the "People who visit URLs" or "People who use apps" variants as your primary signal. They are often too broad. Instead, focus on "People who searched for any of these terms on Google."

Go to your Search campaigns (or your Search Terms report) and find the top 20-30 terms that have actually generated revenue in the last 90 days. Build a custom segment using only these terms. This tells PMax: "The people who buy from me use this specific language. Go find them."

3. In-Market Audiences (Narrow Scope)

The mistake here is "layering for scale." Advertisers add "In-market > Apparel" and "In-market > Shoes" and "In-market > Accessories."

Instead, go as deep into the tree as possible. If you sell high-end running shoes, use "In-market > Sports & Fitness > Sporting Goods > Running Apparel & Shoes." One deep segment is worth ten broad ones. It provides a cleaner "contextual" hint to the algorithm.

The 2 signal types that are noise

In our bench-tests, two signal types consistently failed to move the needle. Including them isn't necessarily "harmful," but it is a waste of time and can lead to a false sense of campaign health.

Similar Audiences

Historically, Similar Audiences were a staple of Google Ads. In the age of PMax, they are redundant. PMax is a "black box" that automatically finds look-alikes based on your conversion data. When you manually add a Similar Audience signal, you are essentially telling the machine to do what it was already going to do. In our 12-account test, adding Similar Audiences resulted in a ROAS shift of ±1.2% - statistically irrelevant.

Demographics

Google's AI is incredibly good at predicting conversion based on behavior (search history, site visits, video consumption). It is less concerned with whether the user is 25 or 45. In the eCommerce space, we've found that PMax often finds high-converting users in demographic buckets that the brand owner never suspected. By applying demographic signals, you may actually be slowing down the discovery of these "hidden" buyer segments.

Signal priority order for new PMax launches

The "Visionary Framework" for a PMax launch isn't about doing everything at once. It's about a staged rollout that allows the algorithm to digest data in manageable bites.

Signal setup walkthrough

Navigating the Google Ads UI to set these up correctly is half the battle. Remember: audience signals are set at the Asset Group level, not the campaign level. This is crucial because it allows you to tailor signals to the specific products or themes within that asset group.

Common signal mistakes

Even with the right strategy, small execution errors can derail performance. Here are the five most common mistakes we see in account audits: