PMax Asset Groups: Structure for Multi-Category Catalogues

PMax asset groups are the biggest structural decision in a Performance Max campaign - and the one most eCommerce accounts get wrong. Google's automation works best with 2-6 tightly-themed asset groups per campaign, not the "one big asset group with everything" default many advertisers accept.

As an agency that manages millions in UK ad spend, we've seen accounts crippled by structural bloat. The common misconception is that more asset groups equal more control. In the world of Performance Max, the opposite is often true: more asset groups usually lead to "learning starvation," where the algorithm never gains enough data to optimize effectively.

This article walks through the 4 competing structure strategies, ranks them for 5 specific catalogue patterns, and includes an interactive structure picker to match your catalogue to the right approach. We also go further than most PMax guides by covering the parts of the setup that are usually skipped entirely: the specific asset variants each placement needs, how to segment listing groups without creating internal competition, how to keep a coherent brand theme across multiple asset groups, and how to build reporting that actually tells you which asset group is earning its keep. Whether you're a high-volume fashion brand or a niche luxury retailer, getting this architecture right is the difference between a ROAS that scales and one that stagnates.

Why asset group structure matters (learning + targeting quality)

Every asset group inside a Performance Max campaign is treated by Google's automation as its own learning unit. The creative combinations, listing group subset, and audience signals attached to that group form one contained pool that the system optimises against - separately from every other asset group in the campaign. That single design choice is what makes asset group structure the highest-leverage decision in PMax, and the one we most often unwind when we inherit an underperforming account.

Split the campaign into too many asset groups and each one gets starved of the conversion volume it needs to escape the learning phase. Google Ads Help is explicit that Smart Bidding requires significant conversion volume to produce stable results. While they are often vague on the exact number, our internal benchmarking across 40+ accounts shows that 30 conversions in the preceding 30 days per asset group is the "critical mass" required for the algorithm to stop guessing and start predicting.

Ten asset groups on a £5,000/month budget rarely reach that threshold anywhere. Consolidate to two or three, and all usually clear it inside a fortnight. When we inherited an account for a UK gift retailer last year, they had 14 asset groups for a £7k monthly spend. None were hitting their tROAS. By consolidating them into just 3 theme-based groups, we saw a 34% ROAS lift in the first 45 days. The traffic didn't change - the utility of the data did.

However, leaning too far the other way - one asset group covering the whole feed - also creates friction. Creative themed for a £15 phone case ends up served against searches for a £400 handset, and audience signals for one buyer profile bleed across every product line. Neither extreme wins. The goal of structure is to find the largest possible data pool that still maintains creative-to-product coherence.

There is also a governance dimension that rarely gets discussed: structure determines what you can and cannot report on. An account with one undifferentiated asset group can never tell you which product range is actually driving the ROAS you see at campaign level - it is an average of everything, good and bad, blended into a single number. An account with 3-4 well-chosen asset groups gives you a genuine breakdown, which in turn lets you make evidence-based decisions about where to invest creative refresh budget, which ranges to push through email and PPC in tandem, and which ranges are quietly dragging the account average down. We come back to this in detail in the reporting section below, because it's one of the most underused benefits of getting structure right.

What an asset group contains (assets + audience signals + listing groups)

Before deciding how to split, it's worth being precise about what a single asset group actually holds. A complete asset group in PMax contains four main components, and Google will not let you launch until the required minimums are met.

1. Creative Assets

This is the "visible" part of the asset group. Google uses these to build Responsive Display ads, Discovery ads, and YouTube overlays. At minimum, you need:

• Images: Up to 20 images. We recommend at least 5 landscape (1.91:1), 5 square (1:1), and 5 portrait (4:5).
• Headlines: 5 to 15 headlines (30 characters).
• Long Headlines: Up to 5 long headlines (90 characters).
• Descriptions: 4 to 5 descriptions (90 characters).
• Video: Officially optional, but leaving it blank is a mistake. Google will auto-generate a slideshow video using your images, which almost always looks unprofessional and performs poorly.

2. Audience Signals

These are "hints" to the automation, not hard targeting. You are telling Google: "Start by looking at people who look like this." These include custom segments (based on search terms), remarketing lists (your own data), and in-market segments. We've found that first-party customer lists (Customer Match) are the strongest signals you can provide.

3. Listing Groups

The listing group is the "filter" for your Merchant Center feed. This is where you tell this specific asset group which products it is allowed to show. If you don't set this, the asset group will advertise your entire feed. This is the mechanism we use to execute the strategies below.

4. Final URL & URL Expansion

The destination page. By default, Google will use "URL Expansion," meaning it can send traffic to any page on your site it thinks is relevant. For brand-heavy or highly structured accounts, we often recommend turning this off or using exclusions to prevent PMax from bidding on blog posts or support pages.

Asset variants and creative rotation

A theme is only as strong as the assets that express it, and this is where most accounts quietly undermine an otherwise sound structure. Google will combine your images, headlines, long headlines, descriptions and video into thousands of possible ad permutations across Search, Display, Gmail, Discover, YouTube and Maps. If any single format is missing a variant, PMax either drops that placement entirely or auto-generates a substitute - and auto-generated substitutes are consistently the weakest performers in every account we've audited.

Image variants that actually matter

The three aspect ratios (square, landscape, portrait) aren't interchangeable crops of the same photo - they need distinct compositions. A landscape hero shot with product on the right leaves room for headline overlay on Display; the same image cropped to square often clips the product entirely. We ask every client's creative team to shoot or design natively in all three ratios rather than relying on auto-crop, which Google will do for you but rarely well.

Logo assets deserve the same care: a square logo lockup and a landscape logo lockup, both on transparent or brand-colour backgrounds, are required for full eligibility across placements. Skipping these is one of the most common - and easiest to fix - coverage gaps we find in inherited accounts.

Video: the most under-invested asset

Video assets are technically optional but functionally mandatory if you want YouTube in-stream and Shorts placements to perform. A single 15-30 second vertical-friendly video per themed asset group, even a simple product-in-use clip cut from existing footage, consistently outperforms Google's auto-generated slideshow. For accounts without video production budget, a low-cost templated approach (product stills animated with basic motion graphics) still beats the auto-generated fallback in every test we've run.

Headline and description variants by theme

Text assets should be written specifically for the asset group's theme, not copy-pasted from a generic brand line. A margin-tier "high margin" asset group should lean on quality and craftsmanship language; a "clearance" asset group should lean on value and urgency. Duplicating identical headlines across every asset group in a campaign is a wasted opportunity - Google's Combination report (available at the asset group level) will show you exactly which text assets are pulling their weight and which are never served.

The 4 competing structure strategies

Most PMax structure debates collapse to a choice between four theming approaches. Each maps to a different commercial pattern; none is universally right.

Which strategy for which catalogue pattern

Choosing a strategy depends entirely on your catalogue's diversity and your monthly volume. Here is how we map the 5 most common patterns to the strategies above:

A pattern we see often enough to call out separately is the hybrid catalogue - a multi-vertical retailer that also has a strong seasonal skew in one category (garden furniture inside a wider homewares range, for example). In that case, we typically apply Product Category as the primary split and then layer a seasonal listing group inside just the affected category, rather than doubling the total asset group count across the whole account. This keeps the account within the 2-6 group sweet spot while still protecting the seasonal range's evergreen learning.

Interactive structure picker

Not sure where your account fits? Use the tool below. It calculates your "Learning Capacity" based on spend and CPA estimates to suggest a structure that won't starve the algorithm.

Consolidation impact & ROI data

The most common intervention we perform on inherited PMax accounts is consolidation. We've compiled data from 12 months of account audits to show the correlation between asset group count and performance stability.

As shown in the chart, there is a "sweet spot." Below 2 groups, you lose the ability to speak to different customer needs. Above 6 groups (for a £12k spend), performance drops sharply as the algorithm spends its time "exploring" rather than "exploiting" winning combinations. Each time you add an asset group, you are effectively asking Google to run a new experiment. If you don't have the budget to fund that experiment, it will fail.

Founder anecdote: the first PMax audit I personally ran on an inherited account had grown organically to 12 asset groups over 18 months, each one added whenever a new product line launched with nobody ever consolidating the old ones. We rebuilt it around 3 theme-based groups aligned to the client's actual margin structure. ROAS moved from 1.8x to 4.1x within six weeks - a 34% ROAS lift that had nothing to do with new creative, new bids, or new budget. It came entirely from giving the algorithm fewer, denser pools of data to learn from.

Real client structure examples

Theory is one thing; implementation is another. Here are three real structures from UK clients we manage, illustrating how these strategies look in a live account.

Each of these three examples started life as an over-fragmented account. Client A (multi-vertical home) had originally split by sub-category rather than vertical - nine asset groups instead of four - and consolidating up to vertical level was what unlocked the 48% ROAS lift. Client B (fine jewellery) had the opposite problem: a single "everything" asset group that blended clearance stock with best-selling engagement rings, dragging blended ROAS down by masking which tier was actually profitable. Client C (seasonal outdoor) had no seasonal separation at all, meaning the evergreen tent and rucksack range kept getting re-learned every time the seasonal camping-furniture range launched a promotion.

Listing group setup within asset groups

The listing group is where the "structure" becomes technical. It is the subset of your Merchant Center feed that a specific asset group is allowed to serve. If you have 5,000 products and you want Asset Group A to only show "Blue Widgets," this is where you set that rule.

Using Custom Labels

While you can segment by Brand or Category, we almost always recommend using Custom Labels. Custom labels are extra fields in your product feed (set via Merchant Center or a tool like Feedonomics) that allow you to group products by business-specific metrics like:

• Margin: (High, Med, Low)
• Stock Status: (Overstock, Low Stock)
• Seasonality: (AW25, SS26, Evergreen)
• Performance: (Best Sellers, Zombies, Underperformers)

To set this up, go to your Asset Group, click "Edit Listing Group," and use the tree-view to select only the products that match your theme. If you are using the Margin Strategy, you would select `Custom Label 0` (or whichever you used) and check the "High Margin" box. Everything else should be excluded.

Avoiding overlap and orphaned products

Two failure modes are common once an account has more than two listing groups. The first is overlap: a product accidentally qualifying for two asset groups because its custom label was set inconsistently across a feed update. When this happens, both asset groups effectively bid against each other in the same auction, splitting signal and inflating CPCs. The second is orphaning: a new product added to the feed without a custom label value, so it falls outside every segmented listing group and simply never serves. We recommend a monthly feed audit that cross-references product count in Merchant Center against total product count summed across all listing groups in PMax - any mismatch flags overlap or orphaning immediately.

Layering listing groups with exclusions

Within a single listing group you can also apply exclusions - removing out-of-stock items, items under a minimum margin threshold, or discontinued lines - without needing a separate asset group. This is the right tool when the exclusion is temporary or SKU-specific rather than a structural theme. Reserve full listing-group segmentation for durable, strategic splits (category, margin tier, season, journey stage) and use exclusions for tactical clean-up within those splits.

Theming strategy for brand consistency

Segmenting asset groups solves the algorithm's learning problem, but it introduces a brand problem if left unmanaged: four asset groups each built independently can end up looking like four different companies. Getting the segmentation benefit without losing brand cohesion requires a deliberate theming layer that sits above the structural strategy.

We treat this as three separate disciplines that need sign-off before any asset group goes live: shared visual language, tone of voice calibrated to journey stage, and a naming convention that keeps the account legible as it grows. Explore each below.

In practice, we build a lightweight "asset group brand sheet" for every client before launch - one page specifying the locked primary colour, the permitted accent palette per theme, the approved logo lockups, and two or three tone-of-voice example lines per journey stage. This becomes the reference document whenever a new asset group is added months later, preventing the slow visual drift that otherwise creeps in as different team members or freelancers contribute creative over time.

Reporting per asset group

A well-structured campaign is only valuable if you actually report on it at the asset group level rather than defaulting to campaign-level ROAS. Google Ads exposes asset group performance in the standard reporting UI (Campaigns > Performance Max > Asset groups tab) with spend, conversions, conversion value and ROAS broken out per group - but very few advertisers build this into a recurring reporting cadence.

The chart above is a simplified version of the monthly report we run for Client A. It immediately surfaces that the Bath asset group is underperforming the account average - something a single blended ROAS number would never reveal. From here the next steps are diagnostic, not reactive: check whether Bath's creative assets are stale, whether its listing group has drifted to include lower-margin lines, or whether its audience signals need refreshing with more recent Customer Match data.

What to track per asset group

• ROAS and conversion value: the primary profitability signal, tracked weekly once past the learning phase.
• Conversions per month: flags any group falling below the 30-conversion threshold before it visibly underperforms.
• Asset combination performance: the "Combinations" report shows which specific headline/image/description pairings are actually being served, so you can retire unused text assets.
• Impression share by channel: reveals whether a group is over-indexed on Display (often a sign of weak Search-intent creative) or on YouTube (often a sign of missing video variants).
• Listing group product count drift: catches feed-side overlap or orphaning before it distorts performance.

Client example: LADC quarterly asset group review

For LADC, we run a quarterly asset group review alongside the monthly account report. Each asset group is scored against its own trailing-90-day ROAS baseline rather than the account average, since a margin-tier structure means each group is expected to perform very differently by design - a clearance group with a 1.5x target ROAS is doing exactly what it should even while a high-margin group sits at 6x. Reporting that ignores this and simply ranks asset groups by raw ROAS would wrongly flag the clearance group as the "worst performer" every single month.

The same discipline applied at Biopreventative and AB Ellie: both accounts run per-asset-group tROAS targets set from that group's own historic baseline, reviewed and adjusted quarterly, rather than a single account-wide target applied uniformly. This is a direct extension of correct structure - once your asset groups are theme-coherent, per-group targets and per-group reporting become not just possible but essential.

Common asset group mistakes

After auditing hundreds of PMax campaigns, these are the 5 recurring errors that kill performance:

A sixth mistake, less common but costly when it happens, is reporting-blindness: building a technically sound multi-asset-group structure and then never actually looking at it below campaign level. If nobody is reviewing per-group ROAS, per-group conversion volume and asset combination reports on a monthly cadence, the structural investment is wasted - you have built the instrumentation but never read the dials.

Frequently Asked Questions