The Only Lever You Have
Let's be direct about what Performance Max takes away from you.
No ad groups. No keyword lists (brand exclusions and search themes aside). No placement targeting at the granular level. No bid adjustments by device, audience, or time of day. Google's algorithm controls all of that.
What you do have: Asset Groups.
Asset Groups are the single structural mechanism through which you can assert any meaningful control over a PMax campaign. If you're not using them strategically — and the majority of advertisers aren't — you've handed full campaign control to Google's automation and you're hoping it figures things out on your behalf.
Sometimes it does. More often, it allocates budget in ways that don't match your business priorities, optimises toward volume rather than value, and gives you no visibility into which products or creative approaches are actually driving performance.
Asset Groups, structured correctly, change that.
What an Asset Group Actually Does
An Asset Group is a container. It defines three things simultaneously:
1. Which products it covers — via Listing Groups. This is the product scope. The Asset Group will only serve ads for the products you connect through Listing Group conditions.
2. What creative assets to use — headlines, descriptions, images, videos, and logos. Google assembles combinations of these assets dynamically to create ads across its inventory (Search, Display, YouTube, Discover, Gmail, Maps). The assets in the Asset Group are what gets shown.
3. What audience signals to provide — customer lists, remarketing audiences, in-market segments. These are hints for the algorithm about who's likely to convert, not hard targeting restrictions.
The algorithm then serves ads using your assets to the audiences it predicts will convert, for the products in your Listing Group. The entire serving decision — channel, audience, ad format, creative combination — is Google's. Your Asset Group is the brief you hand to the algorithm.
The quality of that brief determines how well the algorithm performs.
The Mistake Almost Everyone Makes: One Asset Group for Everything
The default PMax setup is one Asset Group containing all products, a mix of general creative assets, and maybe a customer list. This is what Google's setup wizard encourages. It's also the worst possible structure for anyone running a catalogue with performance variation across products.
Here's the problem: Google cannot differentiate between your Contenders and your Dogs inside a single Asset Group. A Star product — high conversion rate, strong margins, proven demand — and a Dog product — drains budget, converts poorly, wrong price point — both compete for the same budget pool with the same tROAS target.
The algorithm tries to meet your tROAS by shuffling spend across the entire catalogue. Sometimes it does. But you have zero visibility into whether your Contenders are doing the heavy lifting while your Dogs consume a disproportionate share of spend. You can't tell from account-level reporting. You can't adjust by product tier. You can't set different expectations for different parts of your catalogue.
One Asset Group for everything is a blunt instrument. PMax is already a blunt instrument by design — don't make it blunter.
Segment by Performance Tier, Not Product Category
Here's the counterintuitive insight that separates sophisticated PMax operators from everyone else: segment your Asset Groups by performance tier, not by product category.
The instinct is to create Asset Groups by category — one for leather bags, one for canvas bags, one for accessories. This feels logical. It's the wrong framework.
Google already understands product categories natively. Its algorithm knows a leather bag is a leather bag. It knows what queries to show it for, what competing products exist, what a reasonable CTR looks like. You don't need to teach it that.
What Google doesn't know — and what you do — is which of your leather bags is a Star and which is a Dog. That's business intelligence that isn't in your feed unless you put it there.
A Star leather bag and a Dog leather bag are in the same category, but they need completely different treatment:
- Different tROAS expectations (your Star can support aggressive ROAS targets; your Dog will never hit them)
- Different budget priority (Contenders deserve more investment; Dogs should be suppressed or excluded)
- Different creative briefs (Star creative leans into what's proven; Question Mark creative tests new angles)
Segment by performance tier and you can manage each tier deliberately. Segment by category and you're just rearranging deck chairs.
Asset Strength: The Non-Negotiable Foundation
Before getting into structure, understand that Asset Group performance is gated by Asset Strength — Google's score for how well-equipped an Asset Group is to build good ad combinations.
The scoring runs from Poor → Good → Excellent. A Poor-scored Asset Group will have its serving actively limited. Google cannot build enough valid ad combinations to compete effectively across its inventory, so it simply serves less.
Minimum requirements for a competitive Asset Group:
| Asset Type | Minimum | Recommended |
|---|---|---|
| Headlines | 15 unique | 15 (max allowed) |
| Descriptions | 4 | 4 (max allowed) |
| Images | 5 | 10+ (mix of formats) |
| Logos | 1 | 2 |
| Videos | 1 (15–30s) | 2–3 |
A few critical notes:
Headlines must be genuinely unique. "Shop our range" and "Browse our collection" are not unique — they're the same semantic content with different words. Google's system detects this.
Images need format variety. At minimum: one landscape (1.91:1), one square (1:1). Add portrait (4:5) for YouTube. Lifestyle images outperform product-only shots on CTR.
Video is the most commonly neglected asset. If you don't provide a video, Google will auto-generate one from your other assets. Auto-generated videos are consistently low quality. A 15-second well-produced clip — even a simple slideshow with voiceover — will outperform Google's auto-generation every time.
If your Asset Strength is Poor, fix that before worrying about anything else.
Audience Signals: Hints, Not Targeting
A persistent misconception about audience signals: they are not targeting. Setting an audience signal does not restrict your ad serving to that audience.
Audience signals tell the algorithm: "here are people who have historically been valuable to us — use this as a starting point." The algorithm will expand beyond your signals if it finds conversion opportunities elsewhere. In practice, PMax often serves a substantial portion of impressions to audiences outside your signals.
The practical implications:
Do add audience signals. Your customer list is the single most valuable signal — these are people who have already bought from you, which is the strongest conversion intent signal available. Add website visitors, add in-market segments that match your product categories.
Don't add signals that are too narrow. A signal list with 500 people doesn't give the algorithm enough to learn from. Thin signals starve the algorithm rather than guiding it.
Don't expect signals to restrict. If you're seeing impressions from audiences that seem off-brand, that's the algorithm finding conversions you didn't predict. Let it — that's the whole point of PMax.
Listing Groups: The Product Connection Mechanism
Listing Groups are how you connect specific products to a specific Asset Group. They work via conditions — product attributes that you define in your feed. The most powerful mechanism here is custom labels.
Custom labels (custom_label_0 through custom_label_4) are arbitrary text fields in your product feed that you control. They have no effect on organic Shopping results — they exist purely for your own campaign management. This is where you encode your performance tier classification.
Example feed setup:
| Custom Label 0 | Tier |
|---|---|
| Star | Top performers |
| Cash Cow | Steady revenue, moderate margin |
| Question Mark | High potential, underoptimised |
| Dog | Underperforming, low priority |
Then in your Asset Groups:
- Asset Group "Contenders & Cash Cows" → Listing Group condition: custom_label_0 = "Star" OR custom_label_0 = "Cash Cow"
- Asset Group "Question Marks" → Listing Group condition: custom_label_0 = "Question Mark"
- Asset Group "New Arrivals" (if applicable) → custom_label_0 = "New"
- Dogs: excluded via a negative condition on all Asset Groups, or not listed at all
The catch-all "All Products" listing group in each Asset Group will capture anything not covered by your specific conditions. Make sure your catch-all is deliberately assigned to your lowest-priority Asset Group.
A Practical 3-Asset Group Structure for Mid-Size Catalogues
Here's a concrete structure that works for catalogues with 200–5,000 SKUs:
Asset Group 1: Contenders & Cash Cows
Your proven performers. Products with consistent conversion history, acceptable margins, and sufficient data.
- tROAS target: Higher than account average. These products have earned it.
- Budget priority: This Asset Group should receive the majority of your campaign budget.
- Creative brief: Lean into what works. Use proven messaging, strong product imagery, social proof.
- Audience signals: Customer list + past purchasers + high-intent in-market segments.
Asset Group 2: Question Marks
High-potential products that haven't found their footing yet. Could become Contenders with the right optimisation — or could be Dogs in disguise.
- tROAS target: Moderate. Give the algorithm room to explore.
- Budget: Controlled. Enough to generate data, not so much you're haemorrhaging spend.
- Creative brief: Test new angles. If Contenders are running aspirational messaging, test functional/comparison messaging here.
- Audience signals: Broader — include category in-market segments, not just your customer list.
Asset Group 3: New Products / Launches
Only applicable if you're actively launching new SKUs. New products need discovery budget — they have no conversion history, so the algorithm needs licence to explore.
- tROAS target: Lower. Expect lower returns during the discovery phase.
- Budget: Time-bounded. Set a launch budget and review after 4–6 weeks.
- Creative brief: Awareness-focused. You're building signals, not just converting.
Dogs: Do not give them an Asset Group. Exclude them via a negative condition on all groups, or remove them from your feed entirely. Every dollar spent on Dogs is a dollar not spent on Contenders.
When to Split vs When to Consolidate
More Asset Groups is not always better. The algorithm needs conversion volume to learn from. An Asset Group generating 10 conversions per month cannot optimise effectively — there's insufficient signal.
The practical threshold: minimum 50 conversions per 30 days per Asset Group before you split. Below that, consolidate. The performance visibility you gain from splitting is worthless if the algorithm can't learn from the data it's generating.
Corollary: don't split an Asset Group to create "control." You don't get control from structural fragmentation — you get a learning-starved algorithm making random decisions with thin data.
Split when you have genuine differentiation in product tier AND sufficient conversion volume to learn from each tier independently. Consolidate when volume is thin.
Learning Period: What Resets and What Doesn't
Adding a new Asset Group triggers a learning period for that Asset Group specifically. It does not reset the learning period for the whole campaign.
This is important because it means you can iterate on Asset Group structure without necessarily destabilising your entire campaign. Add a new Asset Group for a product launch, let it run, and your Contenders & Cash Cows Asset Group continues operating on its accumulated learning.
What you should avoid: deleting and recreating Asset Groups unnecessarily, or making frequent large-scale changes to asset or Listing Group composition. Each major change requires the algorithm to recalibrate, which means a period of suboptimal performance while it re-learns.
The operational principle: plan your structure before you build it, then make deliberate changes rather than frequent tweaks.
The Bottom Line
Asset Groups are the only strategic instrument inside PMax. Use them to encode your business intelligence — your product tier classification, your creative differentiation, your budget prioritisation — into the campaign structure in a way the algorithm can act on.
The advertisers winning with PMax aren't winning by finding some secret bidding trick. They're winning because they've given Google's algorithm a better brief. Asset Groups, structured by performance tier with proper creative and audience signals, are how that brief gets written.
Set it up right once. Maintain it as your catalogue evolves. That's the job.