Performance Max (PMax) gravitates toward branded queries because they convert easiest, often inflating return on ad spend (ROAS) while increasing the brand tax you pay on demand you already own.. Because PMax gives only thresholded query data, partial channel visibility, and platform-biased attribution, you can’t measure true brand incrementality from campaign reporting alone. You control it with brand exclusions and negative keywords, and you validate it with holdback testing.
In most cases, true incrementality is far lower than reported, so budget decisions should follow measured lift, not platform metrics, even if reported ROAS declines.
Why PMax Absorbs Brand Traffic by Default
PMax prioritizes conversion probability, so it naturally shifts spend toward branded queries (the easiest conversions in the account) regardless of whether they’re incremental.
Because it operates across networks and can compete in your own brand auctions, this often leads to higher costs for demand you would have captured more efficiently through Search.
That behavior isn’t a bug; it’s the algorithm following the path of least resistance. The issue is what it does to cost.
In a standard brand Search campaign, the auction is tightly controlled. Exact matching, highly relevant ads, and strong historical performance drive high Quality Scores (QS), which keeps costs per click (CPCs) low and predictable.
PMax doesn’t stay inside that system.
Instead of asking “what’s the cheapest way to win this branded click,” PMax is asking “what’s the highest-probability path to a conversion across any network?”
That shift introduces two sources of cost inflation.
First, cross-network capture. A user can search your brand, leave the results page, and later convert through a YouTube, Display, or Discovery ad served by PMax. When that happens, PMax takes credit even though your Search campaign could have captured that same user at a lower cost.
Second, internal competition. When PMax is eligible for branded queries, it can enter the same auctions as your brand Search campaign. Even in a second-price auction, that creates a floor, preventing CPCs from settling at the low levels typical of clean brand traffic and pushing costs higher without adding demand.
Instead of paying brand Search CPCs typical for your vertical (often well below open-auction levels), you may be paying higher CPCs due to internal competition or materially more through video and display inventory.
This is difficult to detect at first glance. Google now offers a Channel Performance Report for PMax, which breaks out delivery across Search, Display, YouTube, Discover, Gmail, Maps, and Search Partners. That visibility helps you see where PMax is serving, but it doesn’t tell you whether those impressions and conversions would have happened anyway through your brand Search campaign, and attribution still credits the PMax touchpoint rather than the underlying branded intent.
If total conversions don’t increase, the difference shows up as higher acquisition cost, not growth.
What the Insights Tab Actually Tells You (and What It Hides)
PMax now gives you more search-term visibility than it used to, but the reporting is still incomplete. Search Terms Insights group queries into categories; the newer PMax Search Terms Report exposes some actual queries, subject to thresholds.
That’s enough to identify obvious brand capture, but not enough to measure its cost or prove whether it’s incremental, so quantifying its impact still requires exclusions and controlled testing.
In a standard Search campaign, you’d use the Search terms report to see exactly which queries triggered your ads, what you paid, and how they performed.
PMax now offers two related reports that get you partway there.
Search Terms Insights (Campaign → Insights → Search terms insights) groups queries into clustered Search Categories. Broad labels like:
- “Shoes”
- “Running shoes”
- “Brand + product” (if you’re lucky)
These categories blur intent. A theme like “Shoes” could include brand queries, non-brand queries, or a mix of both. Even when branded demand is present, it’s often hidden inside aggregated groupings.
Search Terms Report (Insights & Reports → Search terms) is the newer addition. It exposes actual search queries that triggered your PMax ads, along with standard performance metrics. The catch is that low-volume queries are suppressed by privacy thresholds, so you don’t see the full long tail. You also can’t see match types, network, or query-level CPCs the way you can in standard Search.
The UI adds friction. Expanding categories in the Insights view requires digging into “View details,” and even the Search Terms Report requires you to navigate into a specific PMax campaign to see its data. Reporting is not real time, which limits same-day analysis.
The result is a partial visibility gap: you can identify obvious branded queries and see that brand is present, but you cannot fully measure how much spend, cost, or conversion volume is tied to it.
You still cannot fully connect:
- match-type behavior at the query level
- CPC at the query level
- which channel served the click
- whether the conversion would have happened anyway
You’re working with better evidence than in 2023, but it’s still not complete measurement.
Insights, Channel Performance, and the Search Terms Report can confirm that PMax is capturing branded demand. They cannot tell you how much it’s costing or whether it’s incremental.
Visibility alone won’t answer the question. You have to control for it and test.
Before running an incrementality test, you need to limit branded exposure inside PMax. Otherwise, you’re measuring a system that is already biased toward capturing existing demand.
How to Add Brand Exclusions to Performance Max
Brand exclusions are the primary lever for limiting PMax’s access to branded queries on Search, Shopping, and YouTube search inventory. Google recommends them over negative keywords for brand avoidance because they automatically catch misspellings, variants, and subsidiary brands that a manual list would miss.
Negative keywords are the supplement, not the substitute. Use them to block specific leakage you spot in the Search Terms Report, while brand exclusions handle the broad coverage.
In PMax, brand exclusions and negative keywords are two different controls that work together.
Brand exclusions suppress traffic that Google classifies as related to a brand on your list. You pick the brand from Google’s predefined list, and the system handles the variant matching under the hood. Google groups queries based on known brand terms, misspellings, and historical behavior. In practice, you’re not managing queries one by one; you’re applying a broad filter that Google maintains for you.
Account-level negative keywords have existed in Google Ads for years. What’s new in 2025 is PMax campaign-level negatives, which rolled out to all advertisers on January 23, 2025, then expanded to the 10,000 cap in March 2025. Account-level negatives are capped at 1,000 and apply across all relevant Search and Shopping inventory in your account, including PMax. Negatives only block on Search and Shopping inventory; they do not affect Display, YouTube non-search, Discover, Gmail, or Maps placements.
Use the two together:
- Apply brand exclusions to every PMax campaign that should not serve on your brand
- Use the PMax Search Terms Report to find specific brand variants slipping through
- Add those variants as campaign-level negative keywords for surgical leakage control
- Use account-level negatives sparingly, since they affect every relevant campaign in the account
This layered control model is effective when PMax is clearly scaling into branded demand.
Brand Lists and Account-Wide Controls
Brand lists live at the account level, but brand exclusions are applied to individual campaigns. The account-wide lever for blocking specific terms is account-level negative keywords, capped at 1,000.
For most accounts, the right baseline is one shared brand list applied as an exclusion to every relevant PMax campaign, plus a small set of account-level negatives for terms that should never trigger ads anywhere.
Path: Campaigns → Settings → select the PMax campaign → Additional settings → Brand exclusions → select the brand list → Save.
You build the brand list once at the account level (Tools → Shared library → Brand lists) and then apply it as an exclusion to each PMax campaign that shouldn’t serve on your brand. There is no single switch that excludes a brand at the account level for all campaigns at once; the application step is per campaign.
For most accounts, that’s the right starting point:
- one shared brand list, maintained centrally
- exclusion applied consistently to every relevant PMax campaign
- minimal ongoing setup as new campaigns launch
Account-level negative keywords (Admin → Account settings → Negative keywords, capped at 1,000) are the closest thing to an account-wide brand control. Use them carefully: they apply to every campaign serving on Search and Shopping inventory, so a poorly chosen negative can suppress brand Search or other campaigns you didn’t mean to touch.
Campaign-Level Brand Exclusions for More Control
Even with a shared brand list applied across PMax, you may want to vary which campaigns the exclusion lands on. Some product lines or regions may benefit from running on brand; others should never.
This is most valuable in complex or enterprise accounts, where a single brand exclusion list applied uniformly is too blunt and different campaigns require different brand strategy.
Path: Select a PMax campaign → Settings → Additional settings → Brand exclusions → Select brands to exclude.
Campaign-level exclusions become important when brand control isn’t uniform across the account.
In more complex setups, you may need to exclude the core brand in some campaigns while still allowing sub-brands or product lines to run, particularly if those are managed as distinct growth levers.
The same applies when separating direct-to-consumer (DTC) and retail strategies, where brand behavior and margin tolerance differ, or when regional variations require different levels of brand protection.
Where Brand Exclusions Break and How to Monitor Leakage
Brand exclusions reduce exposure but don’t capture every variation of how users search for your brand, which means some leakage is inevitable.
Detecting leakage now relies on a mix of direct evidence (the Search Terms Report) and directional signals (Insights, conversion shifts, timing patterns). Together they tell you where the gaps are.
Gaps typically show up in misspellings, abbreviations, sub-brands, and modified queries (e.g., “brand + category”). If your brand isn’t in Google’s list, you also face a manual approval process that typically takes about a week, during which PMax continues running without protection.
As a result, exclusions are not a one-time fix. They require ongoing monitoring to detect where branded demand is still slipping through.
In practice, that means building a simple weekly rhythm around a few signals.
Start with the PMax Search Terms Report. If brand-related queries are still appearing, those are exact terms you can add as negative keywords for surgical control. Pair that with the Insights tab. If brand-related themes start surfacing as “top performing” search categories, that’s an early indicator of broader leakage through gaps in Google’s brand definition.
Next, look at conversion trends. A proportional drop in PMax conversions after exclusions typically signals that brand was a meaningful contributor. If conversions hold steady, that suggests either stronger non-brand performance or continued brand overlap.
Timing patterns add another layer. If PMax and brand Search spike on the same days, especially during promotions, that often indicates shared demand capture.
For deeper validation, some teams use scripts to flag recurring brand variants or track shifts in theme composition over time. These don’t provide full visibility, but they help confirm patterns you’re already seeing.
There are still structural limits you won’t eliminate. Low-volume queries are suppressed, near-brand variations can slip through, and attribution lag can blur cause and effect.
The goal is early detection and directional understanding, so you can quantify leakage and adjust before it materially impacts performance.
Measuring PMax Incrementality With a Holdback Test
Incrementality cannot be measured inside PMax reporting. Only a controlled holdback test can determine whether the campaign is generating net-new demand or simply capturing conversions that would have happened anyway.
By isolating PMax’s impact at the total system level and comparing regions with and without it, you can quantify true lift, evaluate cost per incremental conversion, and make budget decisions based on actual growth, not attributed performance.
Incrementality answers a specific question: would these conversions have happened if PMax were not running?
That’s different from what Google Ads reports, which credits PMax for any conversion it touched regardless of whether it caused it. Because branded users are already highly likely to convert, PMax will consistently get credit for conversions that would have occurred through your brand Search campaign anyway.
A holdback test is how you separate those two.
A geographic holdback test works by intentionally removing PMax from a portion of your traffic and comparing total outcomes against markets where PMax remains active.
Instead of looking at campaign-level metrics, you’re measuring system-level impact.
Designing a Valid Geographic Holdback Test
A valid geographic holdback test isolates PMax by pausing it in a small, representative set of regions while keeping everything else unchanged, allowing you to compare total business performance with and without the campaign.
For the test to be credible, control and test regions must behave similarly before launch, remain structurally unchanged during the test, and run long enough to separate true lift from normal variation like seasonality, regional bias, and conversion lag.
Most teams use states or DMAs (Designated Market Areas, groups of cities like “Miami–Fort Lauderdale” or “Dallas–Fort Worth”) as the unit of split. States are typically sufficient unless you’re operating at very large scale.
In practice, you designate roughly 5–10% of total traffic as your control group, where PMax is paused. Your brand Search and all other campaigns continue running as normal. The remaining regions become the test group, where PMax stays active.
At that point, the structure is set, but this is where most tests fail.
Match the markets, don’t just split them. Before launching, confirm that your control and test regions track each other historically. Pull at least 90 days of pre-period data and check correlation on conversions or revenue. If conversions in both groups trend together over time, the split is usable. If they don’t, the test won’t produce a reliable result. Avoid regions with unusual seasonality, store density, sales coverage, or supply constraints that would push them out of sync with the rest of the account.
Seasonality and regional bias are the biggest failure points. If one group reacts differently to promotions, timing, or demand shifts, what looks like incremental lift may just be normal variation between markets.
If the baseline isn’t aligned, the test is invalid before it even starts.
Power-check before you launch. If your expected lift is small (say, sub-10%), a 5–10% control may not give you enough statistical power to detect it. The smaller the expected effect, the larger the control share you need or the longer you need to run.
Once the test is live, discipline matters more than setup. Freeze the variables you don’t want to test. Do not change budgets, bidding strategies, campaign structure, promo calendars, landing pages, or conversion actions in only one group. The goal is to isolate a single variable: whether PMax is on or off.
Let the test run long enough to stabilize, typically 4–6 weeks. Shorter durations tend to reflect noise from day-of-week variation, promotions, or conversion lag rather than true impact.
Use backend or CRM-validated data when you can. Google Ads conversion reporting carries the same attribution bias you’re trying to test against. If you have GA4 with non-platform attribution, server-side tagging, or a CRM with confirmed orders, use those for the system-level read instead of, or alongside, Google’s reported conversions.
There are a few common failure modes that invalidate results:
- running the test for too short a period
- selecting non-representative regions
- making mid-test changes to bids or budgets
- relying on platform-reported PMax conversions instead of total performance
Most failed incrementality tests don’t fail because of data. They fail because the test design introduced noise.
Reading and Interpreting Results Without Platform Bias
Incrementality is measured by comparing total conversions in regions with and without PMax, not by using platform-reported metrics, since only system-level performance reveals whether PMax is driving new demand.
In most cases, measured lift is significantly lower than reported, so interpreting results requires normalizing for region size and making decisions based on true incremental value, not attributed conversions.
Once the test has run long enough to stabilize, the question is simple: did the regions with PMax produce more total conversions than the regions where it was paused?
This is where most teams go wrong. The instinct is to open Google Ads and look at PMax-reported conversions, attribution paths, or assisted conversions. None of those answer the question you just set up.
PMax will always report conversions it touches, especially in high-intent environments like branded search. That doesn’t mean it caused them.
Instead, you go back to the structure of the test.
You have two groups:
- regions where PMax was active
- regions where PMax was paused
When you evaluate results, you ignore PMax reporting entirely. You pull total conversions and total revenue by location (from Google Ads, GA4, or your backend) and compare performance between those two groups.
Because regions differ in size, you normalize the data using conversion rate, revenue per user, or scaling based on historical share so you’re comparing like-for-like performance.
Once normalized, the read becomes straightforward:
- If regions with PMax produce meaningfully more total conversions, that difference is incremental lift
- If they don’t, PMax is capturing conversions you would have gotten anyway
From there, you quantify:
- how many additional conversions PMax actually generated
- and what you paid for that lift based on PMax spend in those regions
This is the only place where incrementality shows up: at the total system level, not inside the PMax campaign.
Most accounts fall into a narrow range of incremental lift, and that lift is usually lower than platform-reported performance. Platform metrics measure what PMax touched. The test measures what it caused.
As a sanity check: if you pause PMax in control regions and conversions immediately drop off, something is wrong with the test setup. Validate before continuing.
What to Do With the Results
Measured incremental lift should directly determine PMax budget and brand exposure, with low lift driving exclusion and reallocation, and higher lift justifying continued but controlled investment.
Because removing brand often lowers reported ROAS while improving true efficiency, decisions must prioritize system-level performance over platform metrics and be continuously revalidated as PMax behavior evolves.
Use these thresholds as a decision framework, not a universal benchmark:
- <5% lift → restrict or exclude brand and reallocate spend
- 5–15% lift → run PMax in a constrained way and evaluate marginal efficiency
- >15% lift → continue investing, with ongoing optimization and retesting
The complication is that these changes often make reported performance look worse before they improve the business.
Removing brand typically lowers ROAS inside the platform because you’ve taken away the easiest conversions. To stakeholders, this can look like a decline, but in reality, you’re stripping out low-cost attribution and exposing the true cost of incremental acquisition.
This is the tradeoff: reported ROAS goes down, but system-level efficiency (Marketing Efficiency Ratio, or MER, and blended customer acquisition cost, or CAC) improves.
For example, if PMax is spending $50K/month and testing shows ~30% of conversions were tied to brand capture, that implies roughly $15K/month can be redeployed into channels more likely to drive net-new demand.
This is not a one-time decision. As PMax learns, competition shifts, and Google updates the system, the balance between incremental growth and cost inflation will change and needs to be revalidated.
PMax will always find the easiest conversions first. Your job is to decide whether those conversions are actually worth paying for.
