The Performance Max audience signals debate has been going on since PMax launched. Google's official line: "Audience signals help Performance Max find the right customers faster." The agency line: "Audience signals don't really do anything, PMax just goes broad anyway." The truth, like most debates, is in the middle — and varies dramatically by account stage.
I've tested audience signals on six different PMax campaigns across three accounts, with controls. Here's what they actually do, when they matter, and when they're noise.
The TL;DR: Audience signals matter a lot when PMax is cold (first 4-6 weeks, low conversion data). They matter almost nothing once PMax has 50+ conversions and Smart Bidding has settled. The accounts that update audience signals quarterly are wasting effort. The ones that nail them at campaign launch get a meaningful head start.
What audience signals actually are
When you create a Performance Max campaign, Google asks you to provide audience signals. These can include:
- Your own customer lists (Customer Match).
- Website visitors (remarketing lists).
- Custom segments (built from competitor URLs, keywords, app names).
- In-market segments (Google's purchase-intent audiences).
- Affinity segments (Google's interest-based audiences).
- Demographic targets (age, gender, parental status, income).
The crucial wording in Google's documentation: audience signals are suggestions, not restrictions. PMax can and will serve outside your signals if it thinks there's a conversion to be had.
Audience signals don't tell PMax who to target. They tell PMax where to start looking.
What I tested
Three accounts. Six PMax campaigns. All on the same products, same budget, same bidding strategy. The variable was audience signals.
- Account 1 (cold) — brand new PMax launch with zero historical conversion data. Two campaigns: rich audience signals (Customer Match + website visitors + custom segments) vs no audience signals.
- Account 2 (warm) — PMax running for 6 weeks with ~80 conversions. Two campaigns: maintained audience signals vs removed all signals.
- Account 3 (mature) — PMax running for 6+ months with thousands of conversions. Two campaigns: maintained audience signals vs replaced signals with low-quality placeholder (just demographic targeting).
Each test ran for 30 days at matched budget.
What I found
Account 1 (cold launch): audience signals mattered a lot
The campaign with rich audience signals (Customer Match, top website visitors, custom segments built from competitor terms) hit profitable ROAS in 2 weeks.
The campaign without audience signals took 4-5 weeks to find its footing. Total acquired conversions in the 30-day test: 36% fewer than the signals-enabled campaign.
The interpretation: in cold-start PMax, audience signals dramatically shorten the learning period. Smart Bidding starts pointed at a credible audience instead of having to find one from scratch.
Account 2 (warm campaign): audience signals mattered moderately
The campaign that kept its audience signals continued to outperform the campaign with signals removed, but the gap was much smaller — about 8% on ROAS over 30 days.
Smart Bidding had enough conversion data to identify good customers on its own. Audience signals were still helping but adding less marginal value.
Account 3 (mature campaign): audience signals barely mattered
The campaign that swapped good audience signals for placeholder demographic targeting performed within ~2% of the signals-enabled campaign. Statistically borderline.
At this stage, PMax had its own behavioural data on what converts. Audience signals were largely decorative.
When audience signals matter
Based on the testing and similar patterns from other accounts, audience signals matter in roughly this curve:
- First 2-4 weeks of a PMax campaign: very high impact.
- Weeks 4-12: moderate impact, declining.
- Beyond 12 weeks: minimal impact for steady-state campaigns.
Audience signals also matter more when:
- The account has limited historical conversion data (a new account or a new product line).
- The product is niche with non-obvious buyer personas (where Google's defaults are likely off-target).
- You have high-quality first-party audiences (real CRM data, real high-LTV customer lists) — these signals carry significantly more weight than generic in-market segments.
When audience signals don't matter
- Mature campaigns with stable conversion data. PMax has its own behavioural model. Your signals add noise more than insight.
- Broad consumer products where Google's auto-targeting is already accurate.
- Generic audience signals built from in-market or affinity segments without first-party data. These are essentially demographic targeting under a different name.
What to actually put in audience signals
If you're going to use them (and for new campaigns, you should), the priority order:
1. Customer Match — high-LTV customers
Your top-LTV customer list. Refreshed monthly. This is the highest-signal input you can give PMax.
2. Website visitors — converters specifically
Not "all website visitors." Visitors who completed a conversion. Or visitors who made it to step 3 of checkout. The closer to a real conversion, the better the signal.
3. Custom segments — competitor + buyer intent
Build custom segments from:
- URLs of competitor websites that high-intent buyers visit.
- Search terms high-intent buyers use (e.g., comparison terms like "X vs Y").
- Apps your buyers use.
4. Demographic refinement (only if known)
If you have strong evidence about which demographics convert, include them. If not, leave them off — Google will figure it out.
Skip
- Generic in-market segments. They're too broad. They dilute the signal.
- Affinity segments. Even broader. Almost always noise.
- Demographics where you're guessing. Wrong demographic signals actively hurt early learning.
How often to refresh audience signals
Less often than you think:
- At campaign launch: build them carefully.
- At week 4-6: review and add anything you missed.
- Every 3-6 months: refresh Customer Match lists, validate custom segments are still relevant.
- When something major changes: new product line, new target market, new business model.
That's it. Updating audience signals weekly or even monthly is busywork. PMax's own conversion data is the primary signal after the cold-start period.
Use case: a new health supplement brand launching PMax
A composite based on patterns I've seen.
A new direct-to-consumer health supplement brand was launching PMax with zero historical conversion data. The temptation was to skip audience signals — Google would "figure it out" from the products and landing pages.
We invested half a day building proper signals:
- Customer Match: imported the founder's small email list (~2,000 contacts) from her existing newsletter. Low volume but high-intent.
- Website visitors: built a list of people who'd added to cart in the pre-launch soft-open period (~400 visitors).
- Custom segments: built three custom segments based on:
- URLs of competitor supplement brands and review sites. - Search terms like "natural sleep supplement", "magnesium glycinate dosage". - Apps including MyFitnessPal, Headspace.
We then launched the campaign at a modest £80/day budget.
Results over the first 30 days:
- Profitable ROAS hit by day 11.
- 142 conversions in the first 30 days.
- CAC came in 28% below the founder's pre-launch target.
- The campaign was profitable enough by week 3 to triple the budget.
We ran a parallel small-budget control with no audience signals on the same products. It took 23 days to reach profitable ROAS and produced ~60% fewer conversions in the same period.
The audience signals didn't transform the business. They compressed 6 weeks of learning into 2.
Common mistakes
- Skipping audience signals on new campaigns. This is the one stage where they matter most. Don't shortcut.
- Updating audience signals weekly. Busywork. PMax doesn't need it after the first month or two.
- Loading every in-market segment Google offers. Dilutes the signal. Be selective.
- Treating audience signals as restrictions. They aren't. PMax can and will serve outside them. Use exclusions for actual restrictions.
- Ignoring Customer Match. First-party data is the highest-quality signal you can give PMax. Use it.
- Building custom segments lazily. "Type a few keywords" is not a custom segment. Use competitor URLs, app names, and specific intent queries to get real signal.
Bottom line
Audience signals matter most when PMax knows least. They matter least when PMax knows most.
- At campaign launch: invest the half-day to build strong signals. The shortened learning period is worth it.
- In steady state: focus optimisation effort elsewhere (asset groups, negatives, conversion value rules). Audience signals are second-order.
- Quality over quantity: Customer Match + converter remarketing + specific custom segments beats a dozen generic in-market lists.
- Refresh quarterly: don't fiddle weekly.
Most of the "audience signals are useless" claims come from advertisers who only experienced PMax in steady state, where they genuinely don't move the needle much. Most of the "audience signals are critical" claims come from advertisers who only experienced PMax at launch, where they make a real difference.
Both are right. They're just describing different stages of the campaign lifecycle.
Use them right at launch. Stop fussing with them in steady state. Move on to optimisations that actually compound.