Google Shopping and Performance Max only know what your product data tells them. They see a title, a price, a category and an image. They don’t see that one product earns you three times the margin of the next, that another has four units left, or that half the catalog is a winter range and it’s July. Custom labels are the field where you tell them.
Most of the work is deciding what to put in the five labels you get and keeping them true. This post covers how I pick them, how to refresh them without a person doing it every week, how they map to campaign structure in Performance Max and Standard Shopping, and the mistakes that make a label setup worse than none.
What custom labels are, and the rules that shape your design
Merchant Center has five custom label attributes, custom_label_0 to custom_label_4. You define what each one means, and they’re for Performance Max, Shopping and Demand Gen campaigns only. Each product carries one value per attribute, each value runs 1 to 100 characters, and each attribute allows up to 1,000 unique values across the account (Custom label 0–4, as of 2026-11-10). Google’s example is a season label, custom_label_0, with the values winter and summer (Use custom labels for Shopping ads, as of 2026-11-10).
The “one value per attribute” rule is the design constraint. A product can be “high margin” or “low margin” in the same label, never both. So each label holds one question with a few mutually exclusive answers. If you try to cram margin and season into the same label, you end up with values like “high-winter” and “low-summer”, and the combinations multiply fast.
Labels take time: newly created or edited labels can take up to 24 to 48 hours to appear in Google Ads (Use custom labels for Shopping ads, as of 2026-11-10). Plan around that lag when you label a seasonal switch.
Which labels are worth having
Every label has to change a decision you make: a bid, a budget, an include or exclude. If a label only gives you a filter in a report, it takes one of five slots and does nothing else. These are the ones I reach for.
| Label idea | What it answers | What you do with it |
|---|---|---|
| Margin band | How much does this product leave after costs? | Different ROAS targets per band |
| Stock position | Is it well stocked, running low, or overstocked? | Pull back on low stock, push overstock |
| Season | Is it in its selling window? | Include in season, exclude out of it |
| Price band | Low, mid or high ticket | Separate budgets, since order value and conversion rate can differ |
| Performance tier | Best seller, steady, or barely sells | Protect winners, test or exclude the tail |
Margin band. This is the label I set up first. Work out each product’s contribution margin and drop it into three or four bands, then set a target for each. I’ve already written the maths and the trade-offs in breakeven ROAS, so I won’t repeat it here. The label is how that work reaches the campaign.
Stock and sell-through. A product with a couple of units left in a single awkward size can take clicks it has little chance of turning into orders. An overstocked line you want to clear is the opposite case. Both are worth a label, with values like “healthy”, “low” and “clear”. The label only helps if the stock position is current, which I come back to below.
Season. The simplest label and the easiest to forget. Tag products by the window they sell in, then include or exclude whole groups as the calendar moves, instead of rebuilding campaigns each quarter. Because of the 24 to 48 hour lag above, switch a label before the window opens, not on the day.
Price band. Order value affects how much you can afford to pay per click. A cheap accessory and a high-ticket item can behave very differently, and a price band lets you give them separate budgets and targets.
Performance tier. Best sellers go into a campaign where budget is protected. The long tail of products with almost no impressions or clicks gets its own group, so you can see which ones deserve a better title or image and which ones to exclude. I treat this label as a rolling review and not a permanent verdict. A product moves between tiers as data comes in.
Keeping labels fresh automatically
Labels go stale. Stock changes daily, margins move when suppliers change prices, and a “best seller” from last spring may be a dud now. A label you set once and forgot is worse than no label, because the campaign structure keeps acting on it.
There are two routes in Merchant Center, and I often combine them.
Supplemental data source. A supplemental source is a secondary source that adds to or updates details from your primary source. It can’t add or remove products, and it can’t stand alone (Create a product data source, as of 2026-11-10). Google lists file, Google Sheets template and API as the ways to supply one, and it uses the ID to match rows to your products, so the IDs need to line up with your main feed (Create a product data source, as of 2026-11-10). The practical benefit is that your store platform’s main feed stays untouched while the labels live somewhere you can compute and update them. Margin and performance labels in particular often depend on cost and sales data that isn’t in the main feed, so I build them in a sheet or script, fed by cost and sales data, and keep them out of the main feed.
Attribute rules. Google’s page on custom labels says to “use feed rules to assign custom labels automatically” (Custom label 0–4, as of 2026-11-10). The same feature is now called attribute rules, and it transforms your data with conditions that filter items; operations such as “set to” are applied only when the conditions are met (Set up your attribute rules, as of 2026-11-10). Rules suit labels that follow from data already in the feed, like season by category or a label for a specific brand. After you change rules, you need to apply the changes and reprocess the data source (Set up your attribute rules, as of 2026-11-10).
Whichever route you use, a label is only as fresh as whatever updates it. I put a date on every labelling job and check it alongside the feed health review. For stock and performance labels, refresh weekly at least. For margin and season, a refresh when costs or the calendar change is enough.
How labels map to campaign structure
The label does nothing until a campaign splits on it. The two campaign types do this a little differently.
Performance Max. Google recommends using custom labels to group items and then targeting those labels in listing groups. Each asset group can have up to 1,000 listing groups, and Google says a large number (over 1,000) is not a best practice and performance may suffer. To leave a group out, you exclude the listing group or subdivision (Manage a Performance Max campaign with listing groups, as of 2026-11-10). In practice, a label becomes either its own campaign (a high-margin campaign with its own ROAS target and budget) or an exclusion (out-of-season and low-stock products kept out of the main campaign).
Standard Shopping. Product groups start as a single “All products” group, and Google recommends splitting inventory into smaller subdivisions to make bidding more specific. You can subdivide up to 7 levels, each ad group can hold up to 5,000 product groups, and you can’t set a bid on a group that has been subdivided; bids go on the groups without subdivisions (Manage a Shopping campaign with product groups, as of 2026-11-10). So the label becomes a subdivision, and each value gets its own bid. If you’re choosing between the two types, Performance Max vs Standard Shopping covers what you give up with each.
Whichever type you run, keep the number of splits low. The common structure is one label that decides which campaign a product lives in (usually margin), and a second that decides whether it’s in or out (season or stock). Everything else is reporting.
Mistakes I see
Too many splits. Every campaign you carve out gets only its slice of the conversions. Google’s requirement for target ROAS on Search and Shopping campaigns is at least 15 conversions in the past 30 days at the conversion tracking level (About Target ROAS bidding, as of 2026-11-10). If a margin band times a price band times a performance tier leaves a campaign with a few orders a month, the bidding has nothing to learn from. Start with two or three groups, and merge any that run short on data.
Stale labels. Covered above. The signature symptom is a campaign pushing out-of-stock or out-of-season products because the label still says they’re live. Date your labelling jobs and check them.
Duplicating product_type. The product_type attribute is your own categorization, and Google uses the first value to organize bidding and reporting in Shopping campaigns (Product type, as of 2026-11-10). If your categories already express “dresses” and “shoes”, don’t spend a custom label repeating that. Use labels for what the category tree can’t say: margin, stock, season, performance.
Labels without a decision. If you can’t say what you’ll do differently for each value, don’t create the label. A “brand tier” label that never changes a bid is clutter.
Unstable value names. “High” in one run and “high margin” in the next creates two separate values and breaks the listing group that targeted the first. Write the allowed values down and have the labelling job reject anything else.
Using every slot. I leave at least one label free. When a new need turns up mid-year, an empty label saves a restructure.
A starting setup
If you have no labels today, this is where I start. Put a margin band in one label. Put a stock or season flag in a second, used only to exclude. Put a performance tier in a third once you have a few months of data. Build them in a supplemental sheet or script, refresh them on a schedule, and split campaigns on the first label only. Add structure when the account’s conversion volume supports it.
Conclusion
Custom labels are cheap to add and easy to get wrong. Choose labels that change a bid, a budget or an include/exclude decision, keep each label to one question, refresh them automatically, and split campaigns on as few of them as the conversion volume allows.
If you’d like a second pair of eyes on your label schema or your Shopping structure, this is part of my Google Ads for e-commerce work. Product titles and attributes matter just as much for how Google reads your catalog, which I cover in product feed optimization. Schedule a call and bring a feed export.
