An ecommerce manager exports a product performance report from Google Ads for the last quarter and sorts it by impressions. The first few hundred rows look busy. Then the numbers thin out, and somewhere around row two thousand they stop: every product after that point shows zero impressions, zero clicks and zero cost. The store sells twenty-four thousand products. Google Ads has been selling a few hundred of them, and nobody chose that.
Short version: Google Ads for large catalogs is a systems problem, not a campaign problem. Once a store carries ten thousand or more SKUs, nobody can manage products one at a time, so the account has to manage them by rules: a feed built from title templates and complete identifiers, custom labels that tag every product by margin, price band and performance status, a tiered structure that treats heroes, villains, zombies and the untested long tail differently, targets set per margin band instead of one blended number, and negatives and scripts that run at the scale of the catalog. This guide covers each of those pieces, the Merchant Center limits you will meet, a worked margin example and the mistakes that keep big catalogs stuck.
The size of the catalog matters more than the category here. An auto parts store, a home improvement retailer, an industrial supplies site and a general homeware store with tens of thousands of products share more Google Ads problems with each other than with small stores in their own vertical. Our category guides, such as Google Ads for auto parts stores, cover what is specific to each trade. This guide covers what changes when the catalog itself becomes the hard part.
Why Google Ads for large catalogs works
Breadth is a real advertising asset. A catalog with thirty thousand products matches an enormous number of specific searches, many of them with little competition because few stores carry that exact item. Google Shopping and Performance Max can sell that breadth for you when the feed is good: every product works as its own keyword, its own ad and its own landing page, with no one writing ad groups for it.
The economics of the long tail are also friendlier than they look. On its own, an obscure product might sell once a month. Ten thousand obscure products each selling once a month is a business, and specific searches usually cost less per click than the contested category terms. The catch is that this only works as a system. Product-by-product management is impossible at this size, which is why structure and rules beat effort in Google Ads for large catalogs.
Small catalog (a few hundred SKUs)
- You can read every product's numbers each month
- Titles can be written and checked by hand
- One or two Shopping or PMax campaigns are often enough
- Negatives are added by reading search terms line by line
- Problems are visible because the list is short
Large catalog (10,000+ SKUs)
- Nobody reads every product, so status must be labelled by rules
- Titles come from templates per category and supplemental feeds
- Campaigns split by tier and margin band, not by product
- Negatives live in shared lists built from patterns
- Problems hide in averages until a report is sorted the right way
How shoppers search a big catalog
Large catalogs attract a different mix of searches from small ones. A store with a narrow range mostly meets category searches and brand searches. A store with tens of thousands of products meets those plus a long tail of very exact searches, and the long tail is where most of the catalog's value sits.
Part numbers, MPNs and GTINs
In parts, tools, electronics accessories, plumbing, lighting and many trade categories, people search for the exact code printed on the old product or in the manual: a manufacturer part number, a model number or a replacement code. These searches carry very high intent, because the shopper already knows what they need and is choosing where to buy. Google can only match your product to them if the identifier is in the feed. That means the MPN attribute filled correctly, the GTIN where the product has one, and for many categories the part number written into the title as well.
Model plus attribute searches
The next layer is model or product type plus an attribute: a size, a color, a material, a compatibility, a pack count or a voltage. "Brass cabinet hinge 35mm soft close", "replacement filter for [model]", "LED bulb E27 warm white 6 pack". Each attribute a shopper adds narrows the set of products that should show. A title that contains those attributes enters the search; a title that holds only a product name does not. On a large catalog, the gap between those two kinds of title adds up across thousands of products.
Category head terms
Broad category searches, such as "kitchen faucets" or "office chairs", have the most volume and the most competition, which is why our guide to Google Ads for home and kitchen brands leans on spec-level product terms instead. Large catalogs can win some of them, but usually on the strength of a proven set of best sellers rather than the whole range. In most accounts these searches belong to the hero tier and to a small Search campaign on proven categories, not to the long-tail pool.
The zero-impression problem
Pull a report of impressions by product for the last 90 days and count the products with none. In unmanaged large catalogs that number is often a large share of the approved feed. These products are in Merchant Center, approved and eligible, and yet invisible. In Google Ads for large catalogs, this is often the single most revealing number in the account. Our Google Shopping audit checklist treats this count as one of the first findings to write down.
There are three usual causes, and they need different fixes:
- Weak feed data. Titles without product type or attributes, missing identifiers and missing categories give Google too little to match. The fix is in the feed.
- Budget concentration. Smart Bidding favors products with conversion history, so early winners keep getting served and new products never collect the data they would need to compete. The fix is structure.
- No real demand. Some products simply are not searched for. The fix is to accept it and stop letting them sit in campaigns that matter.
Telling these apart is the work. A product with a poor title in a category that sells well is a feed problem. A product with a strong title and solid demand stuck behind best sellers is a structure problem. A product nobody searches for is a catalog decision, not an advertising one.
Feed operations for Google Ads for large catalogs
With thousands of SKUs, nobody is writing keywords. The feed does that job, and at this scale feed quality is an operations question: rules and templates applied across categories, not hand-polishing individual items. The general method is in our guide to Google Shopping feed optimization, and the case for treating the feed as the main lever is in the product feed most accounts ignore. These are the parts that matter most when the catalog is large.
Title templates by category
Write one title pattern per category, in the order shoppers search that category, and generate titles from product data. A hinge category might use type, size, finish, feature and pack count. A filter category might use product type, compatible model, part number and brand. Google allows titles of up to 150 characters and shows only the first part in most placements, so put the most important words first and use the full length for matching. A title such as "Folding Blackout Door Curtain, Pleated Accordion Panel with Hooks, 36 x 80 in, Grey" enters dozens of real searches. A code-led title enters almost none.
Identifiers: GTIN, MPN and brand
Fill GTIN wherever a product has one, MPN and brand wherever the GTIN is missing, and keep all three consistent with each other. Large catalogs often inherit bad identifiers from supplier files: one GTIN copied across all variants, internal SKUs typed into the GTIN field to clear an error, brand names spelled three different ways. Wrong identifiers are worse than missing ones, because they can tie your product to someone else's listing or trigger disapprovals. Our guide to Google Shopping feed errors covers the identifier issues in detail.
Variants and item_group_id
Products that come in sizes, colors or pack counts should be submitted as variants that share an item_group_id, with each variant carrying its own distinguishing attributes, image and, ideally, a title that names the variant. Google's guidance is to put the variant detail, such as color or size, in each variant's title. Done properly, each variant can show for its own searches. Done badly, one variant's image and price stand in for all of them, and the shopper who searched for a specific size lands on the wrong one.
Supplemental feeds and feed rules
The platform export will rarely produce everything Merchant Center needs. Two tools fix that without touching the store's product data. Feed rules transform attributes in bulk, for example building titles from several fields or mapping your categories to product types. Supplemental data sources, often a Google Sheet, add or override attributes for products that already exist in the primary feed, matched by ID. A supplemental source cannot add new products, so it is the right home for custom labels, corrected titles and missing identifiers, not for listing items the primary feed leaves out. The IDs must match exactly, or the row changes nothing.
Custom labels: the five you get
Merchant Center gives you five custom label attributes, custom_label_0 to custom_label_4. Each label can hold up to 1,000 unique values across the account, and values over that limit are not used for reporting or bidding. Five labels is not many for a catalog of thirty thousand products, so plan them before you fill them.
A custom label plan that works for most large catalogs
custom_label_0: margin band (high, medium, thin). custom_label_1: price band (for example under 25, 25 to 100, 100 to 300, over 300). custom_label_2: performance status (hero, villain, zombie, new, long tail), updated by a script or a scheduled sheet. custom_label_3: season (spring, summer, autumn, winter, all year). custom_label_4: kept free for promotions, clearance or a test. Keep the values short and few, so the labels stay well inside the 1,000 value limit and stay readable in reports.
Merchant Center limits and disapprovals at scale
Merchant Center accounts have a product quota. Google's quota help page lists a default of 150,000 items for a standard account, and you can request more if the account uses most of its quota, performs well and keeps its disapproval rate low. Very large catalogs, especially with many variants and several countries, can reach it, and products over the limit are simply not submitted.
One feed error can remove hundreds of products at once
On a small store, a disapproval affects a few items you notice. On a large catalog, a single broken rule, a price that mismatches after a sale starts, or a shipping setting missing for one country can disapprove hundreds or thousands of products in one refresh, and the campaigns keep running on whatever is left. Check the item issues list in Merchant Center every week, watch the disapproval rate rather than the count, and fix causes at the source rather than item by item.
Campaign structure for Google Ads for large catalogs
The structure that works manages tiers, not products. Each tier gets a budget and a target that suits how much is known about the products in it, and products move between tiers on rules.
| Campaign | Purpose | Why it is separate |
|---|---|---|
| Brand Search | Isolate the traffic you already own | Keeps every other number honest |
| PMax (hero tier) | Proven sellers with strong margin | Deserve aggressive budget at a confident target |
| PMax or Shopping (workhorse tier) | Steady sellers in the middle | Managed at a standard target, reviewed monthly |
| Long-tail discovery pool | Everything unproven, at low priority | Lets products earn promotion without real budget risk |
Brand search goes first and stays separate, with brand exclusions on Performance Max so it does not take credit for demand that Brand Search already captures. After that, the tiers are defined by the performance status label in the feed. For how Performance Max and Standard Shopping fit together in this kind of setup, see PMax vs Standard Shopping vs Demand Gen.
Heroes
Heroes are the products with a sustained record of sales at a return above their break-even. They get their own campaign, enough budget that it is not limited most days, and a target set from their margin band. Splitting heroes further by margin band is worth it once each split still has enough conversions for Smart Bidding to work with.
Villains
Villains are products that take a meaningful share of clicks and spend but return well below break-even. They are often popular products with thin margins, or products whose price is not competitive, a pattern covered in depth in our guide to Google Ads for electronics stores. Do not leave them in the hero campaign, where they drag down the target. Move them into a campaign with a stricter target, fix the price or the page, or exclude them.
Zombies
Zombies are products with no impressions, or almost none, over a long window. They do little harm by themselves, but they are a signal. Some are feed problems that a better title or identifier would fix. Some are products with no demand. A quarterly zombie review that sorts one group from the other is one of the most useful maintenance tasks in a large catalog.
The long-tail pool
Everything that is new or unproven goes into a discovery campaign with a modest budget and a lower bar, so products get a fair chance to collect data without risking the budget of the proven ones. This is where Standard Shopping can still be useful, because it gives you product group bids and full control over what is in the pool. Products that sell here move up; products that collect clicks without sales move out.
The monthly operating rhythm
- Refresh the performance labels. Pull product-level data for a window long enough to judge, often 60 to 90 days, and update the status label for every product from fixed rules.
- Promote and demote. Products that sold in the long-tail pool above a set return move to the workhorse or hero tier. Products that spent past a cost limit without a sale move to the villain group or out.
- Review the zombie list. Check a sample of zero-impression products for fixable feed problems, then fix the cause at the template level so the fix covers the whole category.
- Check feed health. Read the item issues list, the disapproval rate and the approved product count against the products you actually sell.
- Read search terms by pattern. Add negatives to shared lists and note categories whose searches suggest a Search campaign is worth building.
Done every month, this turns the account into a way of finding winners in the catalog instead of a bet on guessing them in advance. Our honest view of what Performance Max does with this structure is in the truth about Performance Max for e-commerce.