Scale is supposed to make advertising easier.
But there’s a version of the Google Ads problem that only shows up past a certain catalogue size. It’s not about bid strategy. It’s not about campaign structure in the conventional sense. It’s about what happens to data quality, signal strength, and margin distribution when you spread a finite budget across hundreds or thousands of products.
Small e-commerce brands have a simpler problem. Their catalogue is tight, their top SKUs are obvious, and their conversion data concentrates quickly. The algorithm gets a clean signal. Optimisation follows.
Scale that to 500, 1,000, or 5,000 SKUs and the same account architecture produces a fundamentally different outcome. The tools scaled. The strategy didn’t. And the gap between the two is quietly compounding in your P&L.
The Scale Paradox
A larger catalogue means more opportunity, more products to match against more search queries, more coverage, and more chances to convert. This is true in theory.
In practice, it creates three interconnected problems that don’t exist at a smaller scale, and that Google Ads’ default settings are specifically not designed to solve.
The larger the catalogue, the larger the decision tree the algorithm has to navigate:
Which products deserve impressions?
Which products should receive budget?
Which products should enter auctions?
Which search queries map to which SKUs?
Which products deserve remarketing exposure?
Which products should receive high-intent traffic?
Here are the top three problems:
Feed quality degrades at the edges
Your top 50 SKUs have optimised titles, clean images, accurate GTINs, and product types that map correctly to how people search. SKUs 300 to 600 have whatever the product team uploaded when the catalogue launched. PMAX treats them identically.
Conversion data thins out across the long tail
Smart Bidding needs volume to work. With budget spread across hundreds of products, most SKUs never accumulate enough conversion data for the algorithm to bid with any confidence. You end up with it guessing on 80% of your catalogue and charging you for the guesses.
Margin distribution is rarely uniform
A 500-SKU catalogue almost certainly spans margins from 8% to 55%. A single blended tROAS target is simultaneously too aggressive for your most profitable products and dangerously lenient for your worst. One number cannot do what five numbers could.
Each of these problems is manageable in isolation.
The issue is that PMAX, the campaign type now handling the majority of Google Shopping spend, compounds all three simultaneously.
What PMAX Actually Does At Scale
Performance Max is an optimisation engine. It takes your asset groups, your budget, your tROAS target, and your conversion signals, and it allocates spend toward whatever it believes will hit your target most reliably. In a small catalogue, this works reasonably well. In a large catalogue, the mechanics work against you.
PMAX is drawn to confidence. It allocates budget toward products with strong conversion histories, consistent click patterns, and reliable revenue signals. These products get more impressions, accumulate more data, and attract even more budget. The flywheel spins.
The products that don’t make it onto the flywheel? They sit in the long tail, receiving occasional impressions and thin data. Their conversion rates look poor — not because the products are poor, but because they've never had enough traffic to prove themselves. PMAX loses confidence and withdraws further. The data gets thinner. The cycle repeats.
The products PMAX is most confident about are rarely your most profitable ones. They're your most predictable ones. Those two things are not the same.
Now layer in the margin problem. The products with strong conversion histories, the ones PMAX favours, tend to be high-volume, frequently purchased, competitively priced items. In most catalogues, these are exactly the SKUs with the thinnest margins. Easy to buy, easy to convert, easy to lose money on at scale.
The result is a predictable pattern: the budget concentrates on low-margin, high-volume products. High-margin products with lower conversion frequency are starved of impression share. The blended ROAS looks respectable.
The margin profile deteriorates quarter by quarter.
- ~20% of SKUs typically receive 80%+ of PMAX spend in large catalogues
- 6–8x break-even ROAS required for a 15% margin product and most blended targets miss it
- 47pt typical margin spread between best and worst-margin SKUs in a 500+ catalogue
The Feed Quality Problem Nobody Talks About
The conversion signal problem gets attention. The feed quality problem doesn’t, and it’s doing just as much damage.
A product feed at scale is almost always a two-tier system, even if nobody designed it that way. Hero products (the ones that drive most revenue) receive care and attention. Titles are written with search intent in mind. Images are shot properly. Categories are accurate. These products match well against relevant queries and convert accordingly.
The rest of the catalogue is typically populated from an ERP or PIM system with minimal intervention. Titles are manufacturer descriptions. Images are stock photos on white backgrounds. Product types are generic. These listings match poorly against intent, generate low CTR, accumulate thin data, and confirm PMAX’s instinct to deprioritise them.
The Compounding Effect
Poor feed quality reduces CTR. Low CTR reduces data volume. Thin data increases Smart Bidding uncertainty. Higher uncertainty pushes PMAX toward safer, better-signalled products. Those safer products receive more budget, accumulate more data, and look even more attractive by comparison. The long tail doesn’t fail. It never gets the chance to succeed.
For a 50-SKU brand, a feed audit is a weekend’s work.
For a 1,000-SKU brand, it’s a structural problem, one that most agencies never address because it sits outside the campaign management remit.
Why Enterprise eCommerce Teams Feel “Blind”
The biggest frustration large brands have with Google Ads today is not performance.
It’s visibility.
Teams often cannot clearly answer:
- Why spend shifted
- Why a category dropped
- Why one SKU scaled
- Why another disappeared
- Which search intent changed
- Whether growth is incremental
- Whether Performance Max cannibalised existing demand
That uncertainty creates strategic risk.
Especially for brands spending:
- $100K/month
- $500K/month
- $1M+/month
At that level, even small allocation inefficiencies create massive wasted spend.
And because automated systems abstract decision-making away from humans, diagnosing problems becomes slower and more difficult.
What The Data Actually Looks Like
Here’s the pattern you’ll find when you join PMAX spend data to your margin feed. The specific numbers vary by sector, but the shape is consistent across large catalogues.
High-volume accessories & consumables
- Margin range: 9–16%
- Break-even ROAS: 6.3x–11.1x
- Actual PMAX ROAS: 3.8x
- Share of spend: 42%
- Share of SKUs: 31%
- Status: Loss-making
Mid-range core products
- Margin range: 22–30%
- Break-even ROAS: 3.3x–4.5x
- Actual PMAX ROAS: 4.1x
- Share of spend: 39%
- Share of SKUs: 44%
- Status: Marginal
Premium & professional range
- Margin range: 38–52%
- Break-even ROAS: 1.9x–2.6x
- Actual PMAX ROAS: 3.2x
- Share of spend: 19%
- Share of SKUs: 25%
- Status: Profitable
In this example, 42% of the PMAX budget is allocated to products that cannot break even at the account's tROAS target. The most profitable segment, where break-even ROAS is well below what PMAX is actually achieving, receives less than a fifth of total spend.
The blended ROAS across all three tiers? 3.9x. Which looks reasonable in the monthly report. This is why this pattern persists for months or years without intervention.
What Good Looks Like At 500+ SKUs
Fixing this isn’t a campaign rebuild. It’s a series of structural decisions that change the information PMAX is working with, and the guardrails within which it operates.
Segment PMAX campaigns by margin tier, not by product category
The default instinct is to organise campaigns by brand or product type. That's the wrong axis. Organise by margin band. Your high-margin products need a lower tROAS target because they're profitable at a lower return. Your low-margin products need a higher target, or they shouldn't be in PMAX at all. One campaign with one tROAS target cannot serve both.
Use custom labels to enforce margin segmentation at feed level
Custom labels are the mechanism that makes margin-tier campaigns possible. Assign each SKU a label that reflects its margin band, whether that's three tiers or five, and use those labels to control which products enter which PMAX campaign. Review and update quarterly as margins shift. This is not a one-time task; it's an ongoing data process.
Pass margin-adjusted conversion values back to Google Ads
If you can send conversion values that reflect profit rather than revenue — even as an approximation by margin tier — PMAX's optimisation objective changes fundamentally. It stops maximising for revenue and starts approximating for profit. This requires technical implementation but produces the most durable structural improvement of any intervention on this list.
Triage your feed by performance tier, not alphabetical order
Audit your product feed in the order that the data tells you to. Start with the SKUs receiving the most PMAX spend. These are the products where feed quality has the highest leverage — a better title or more accurate product type directly affects how well PMAX matches them against intent. Work downward from there rather than attempting a full catalogue overhaul at once.
Reconsider which long-tail SKUs belong in PMAX at all
Products with consistently thin conversion data and low margins are not PMAX candidates. They dilute your campaign's signal quality and absorb budget that your profitable SKUs could be using. Move them to Standard Shopping where you control the bid, or exclude them entirely until you've resolved the margin or feed issue that's making them unviable.
The Conversation Most Agencies Avoid
Here is why this problem persists in most large-catalogue accounts: addressing it makes the headline ROAS number worse before it gets better.
When you pull loss-making SKUs out of PMAX, revenue falls. Conversions fall. The blended ROAS may actually improve because you’ve removed the products dragging it down, but total spend and total revenue both decline.
In an environment where account performance is reported monthly against revenue targets, that's a difficult presentation to make.
The reframe is straightforward, but we need to make it explicit: we are not optimising for revenue. We are optimising for profit. Those two objectives produce different decisions. A business that needs to fund growth from its gross margin cannot afford to fund it from products that don't have any.
The metric that tells the real story isn’t ROAS. It is the contribution margin generated per pound of ad spend across the full catalogue, tracked over time, broken down by margin tier. That number will initially look worse than the ROAS report. Within two quarters, in most accounts, it won’t.
The Google Ads challenge for large-catalogue ecommerce brands is not a campaign management problem. It is a data infrastructure problem — the quality of what you feed the algorithm, the specificity of the signals you give it, and the rigour of the guardrails you build around it
PMAX will optimise toward whatever you’ve told it to. At 500 SKUs, the cost of telling it the wrong thing is proportionally larger. So is the upside of getting it right.
Need a fresh perspective? Let’s talk.
At 360 OM, we specialise in helping businesses take their marketing efforts to the next level. Our team stays on top of industry trends, uses data-informed decisions to maximise your ROI, and provides full transparency through comprehensive reports.
.png)
