Custom labels vs automated product bucketing: which approach fits your catalogue

Custom labels are five open attributes in your Google Merchant Center account that let you tag every product with a value you define, then group products in Google Ads by that value. Automated product bucketing is the same idea applied by software. A tool reads live performance data, sorts each item into a segment, and writes the result back to the custom label field on a schedule. The choice between the two comes down to catalogue size and how fast your product data changes. Both methods help you segment your products and optimize spend across different product groups.

This guide covers both approaches, the methods that sit between them, and the point at which one stops working and the other starts. It also covers when to use feed rules and when a tool earns its place in the business.

‍

Which tool is best for uploading item data to ad platforms

There is no single best tool. The right option depends on what you are uploading and how often it changes.

For static product data such as title, price, availability and product type, a feed management tool like Channable, DataFeedWatch or Feedonomics is the standard choice. These tools connect to your store, transform the data source, and push a clean Google Shopping feed to Google Merchant Center and other ad platforms. They handle mapping, rules and multi channel distribution.

For performance data such as ROAS, clicks and margin, feed management tools are the wrong layer. This is the category people mean when they search for a labelizer. They read your catalog. They do not read your Google Ads account, so they cannot see product performance. To set custom labels based on how a product actually performs, you need a tool that pulls campaign performance from Google Ads, calculates a segment, and writes that value back into the feed. Clarmix and ProductHero both sit in this category.

For small catalogues, Google Merchant Center itself is enough. Feed rules and supplemental sources are free, native, and require no third party tool at all.

The short version. Use a feed manager to get product data in. Use a performance tool to keep custom labels current. Use Merchant Center alone if you have fewer than a few hundred SKUs.

‍

What custom labels are in google shopping

Custom labels are numbered custom_label_0 through custom_label_4. You get five per product. Each account supports up to 1,000 unique values for each custom label attribute, with a ceiling of 5,000 labels in total across all five. Each product can carry only one value per attribute.

You choose what each label means. Google does not define them, and there is no recommended set. That is the whole point of the attribute. One advertiser might set custom_label_0 to season, another to margin band, another to stock age.

Two things are worth understanding before you start. First, custom labels have no effect on search matching. They do not influence which queries your Shopping ads appear for, and they do not appear to shoppers. They exist for segmentation and reporting inside your campaign structure, which is what makes custom labels for Google Shopping different from attributes that drive matching. Second, the attribute is supported in Performance Max, Shopping and Demand Gen campaigns only. For Display with dynamic remarketing you need different attributes.

The value is in what custom labels allow you to do afterwards. Assigning custom labels to every item takes minutes to plan and pays off later. Once every item carries a label, you can build listing groups around that label, split budget between different campaigns, set a separate bid strategy for each group, and read performance by segment rather than by product category.

‍

Why product feed segmentation matters for campaign performance

Most Google Shopping campaigns treat every product the same. The same bid strategy applies to a bestseller and to an item that has not converted in six months. The same budget covers a high margin product and a loss leader.

That is fine at low spend. It stops being fine once your catalogue passes a few hundred items, because averages hide the detail. A campaign showing a 4x return can contain one group of products at 12x and another burning budget at 0.5x. The blended number looks acceptable. The underlying distribution is not.

Segmentation fixes this by making the distribution visible. When you group products by margin, by season, or by performance tier, you can see which segment earns the spend and which does not. Then you can act. Raise target ROAS on the weak group, lower it on the strong one, or exclude the dead stock entirely. High margin products and low margin products stop being treated as one pool.

Custom labels are the field you use to make that segmentation possible in Google Ads. Everything below is a different way of getting values into that field.

‍

Method one: setting custom labels manually in your product feed

The simplest option. Creating a custom label this way means you add a column to your product data file, name it custom_label_0, and fill in a value for every specific product.

This works if your source file is something you control directly. Many e-commerce platforms let you add a custom field per product and map it into the feed. You set the field once, and the value flows through on every upload.

Adding custom labels to your product feed by hand suits stable attributes. Season is a good example. A winter coat stays a winter product. A summer dress stays a summer product. You set the value once and revisit it twice a year.

It also suits small catalogues. With 50 or 200 SKUs, filling a column by hand takes an afternoon. With 20,000, it is not a real option.

The limit is maintenance. Manual values do not update themselves. The moment your label depends on something that moves, such as stock level, price competitiveness or sales performance, manual entry becomes a recurring job that nobody does.

‍

Method two: spreadsheet workflows for custom labels

The next step up. You export your product catalog to a spreadsheet, calculate label values with formulas, and upload the sheet back to Merchant Center as a data source.

This is a genuine improvement over manual entry because the logic lives in the formula rather than in your memory. A margin band can be calculated from cost and price columns. A price tier can be derived with a nested IF statement. Anyone on the team can read the rule and understand why an item landed in a specific group. This helps when several people touch the account.

Google Sheets works well here because Merchant Center can fetch directly from a sheet on a schedule. Update the sheet and the values update in your account without a manual re-upload.

The weakness is data freshness. A spreadsheet only knows what you paste into it. To label products by performance you would need to export a Google Ads report, paste it into the sheet, recalculate, and repeat. Weekly at best. Realistically monthly, and then not at all after the third month.

Spreadsheets are good for logic that depends on catalog attributes you already have. They are poor for logic that depends on advertising results.

‍

Method three: using feed rules in google merchant center

Feed rules are the native automation layer. You can use them to assign custom labels automatically based on values you already submitted in your product data, such as applying a value to custom_label_0 according to price range.

The setup is straightforward when you use feed rules. In your data source, open the attribute rules, add a rule targeting the custom label field, apply your conditions, and set the output value. A rule might read: if price is greater than 100, set custom_label_0 to high value.

Feed rules have real advantages. They are free. They live inside Merchant Center, so there is no extra tool in the stack. They run on every feed refresh, which means values stay current with your catalog automatically. And they scale to any catalogue size, because the rule does not care whether it evaluates 500 items or 500,000.

They also have a hard ceiling. Feed rules can only reference data that is already in the feed. Price, product type, brand, availability, custom fields you supplied. Nothing else. Google Ads performance is not in the feed, so no feed rule can segment by ROAS, clicks, conversions or cost.

This is the single most important limitation to understand. Every method up to this point can only sort products by what they are. None of them can sort products by how they perform, which limits how far you can optimize on results alone.

‍

Method four: supplemental feed labelling

A supplemental source is a secondary data file that updates existing products in your primary feed. You match on product ID and overwrite or add specific fields.

For custom labels this is a clean pattern. Your primary feed stays untouched and owned by your platform or feed manager. A separate file carries nothing but product ID and label values. Update the supplemental file and the labels change. Nothing else in your product data is at risk, which is the safest way to add custom labels in your product data when a feed manager owns the primary source.

Supplemental sources are also how most external tools write labels back. The tool cannot edit your primary feed, so it publishes a supplemental file and Merchant Center merges it in.

This method solves the ownership problem. It does not solve the calculation problem. Something still has to decide what value each item gets. A supplemental feed is a delivery mechanism, not a decision engine. If you build it manually from a spreadsheet, you inherit every freshness issue from method two.

‍

Method five: automated product bucketing

Automated bucketing closes the gap. A tool connects to both your Google Ads account and your Merchant Center account. It reads performance for every item, applies a classification rule, and writes the resulting segment into a custom label field through a supplemental source. It repeats on a schedule, usually daily.

ProductHero popularised this pattern with the Labelizer, which classifies products on two axes, clicks and return on ad spend, and assigns one of four buckets. The name stuck, and labelizer is now used loosely to describe any tool that writes performance based custom labels back to Merchant Center. Clarmix uses six segments including Profitable, Costly and Zombies, and allows the underlying calculation and thresholds to be changed.

The mechanical difference from every earlier method is the data source. Feed rules read the feed. Bucketing tools read the campaign. That is why bucketing can answer questions the other methods cannot, such as which products consumed budget without converting, which ones convert but are held back by low impression share, and which have gone quiet.

The second difference is cadence. A manual process updates when someone remembers. An automated process updates every day, which means your product groups reflect last week rather than last quarter.

The trade-off is cost and dependency. You are adding a tool, a subscription and an integration. For a catalogue of 200 products with modest spend, that overhead is hard to justify. For 20,000 products, the manual alternative does not exist.

If you are comparing options in this category, we broke down the main ProductHero alternatives for Google Shopping management including where each one fits.

‍

Common use cases for custom labels

These use cases apply regardless of which method you pick, and most accounts include two or three of them.

Seasonality is the classic example. Set custom_label_0 to season with values such as Summer, Winter, spring and autumn. Then run separate campaigns per season, or simply exclude out of season items so they stop consuming budget in July.

Margin is the most commercially useful. Group products into high, medium and low bands so margin products can carry their own targets. A 6x return on a 15 percent margin item and a 6x return on a 60 percent margin item are not the same outcome, and your bid strategy should reflect that.

Sales performance covers bestseller and clearance tagging. Bestsellers deserve aggressive bids and their own budget. Clearance stock deserves visibility while it lasts and nothing after.

Price tier lets you separate cheap accessories from expensive core products. Conversion behaviour differs sharply across price bands, and a single target ROAS across both will underperform on one of them.

Stock level protects your spend. Label items that are running low, then reduce or pause bidding before you sell out and pay for clicks on an unavailable listing.

Product age separates new arrivals from established items. New products have no conversion history, so they need a different bidding approach and a period of protected budget to gather data.

Supplier or brand matters when your agreements differ. If one brand carries better terms, that difference should be visible in your campaign structure.

‍

Catalogue size and SKU velocity: where the decision actually sits

Two variables decide which approach fits.

The first is catalogue size. Under roughly 500 products, manual and spreadsheet methods remain workable. Between 500 and 5,000, feed rules become the sensible base layer. Above 5,000, any method that requires a human to review individual items has already failed.

The second is SKU velocity, meaning how fast your catalog and its performance change. A furniture retailer with 300 stable products has low velocity. A fashion retailer cycling collections every six weeks has high velocity. Velocity matters more than raw size, because it determines how quickly a label becomes wrong across different product segments.

Put the two together.

Small catalogue, low velocity: manual labels or a simple spreadsheet. Adding a tool here solves a problem you do not have.

Small catalogue, high velocity: feed rules. Free, native, and current with every refresh.

Large catalogue, low velocity: feed rules for structural attributes, plus periodic manual review of the segments that carry the most spend.

Large catalogue, high velocity: automated bucketing. Nothing else keeps pace.

There is also a signal independent of size. If the segmentation you want depends on performance rather than product attributes, no feed rule can produce it. That requirement alone points to a tool, whatever your SKU count.

‍

Best practices for custom labels in google shopping campaigns

Document what each label means before you set anything. Five unlabelled fields become unreadable within months, particularly when a client account passes between people. Write down that custom_label_0 is season and custom_label_2 is margin band, and keep it somewhere the whole business can see. We recommend a shared sheet rather than a personal note.

Reserve one label for testing. Four for production, one for experiments. It saves you from rebuilding a live campaign structure to try an idea.

Keep values coarse. Three to six values per label is the useful range. Twenty values produce product groups too small to gather meaningful data, and the segmentation stops informing anything.

Do not encode data that already exists as its own attribute. Brand, product type and product categories are separate fields you can already segment on. Spending a custom label on them wastes one of five slots.

Apply labels consistently across the account. A product that is missing a value falls outside every listing group built on that label, which usually means it stops serving without anyone noticing.

Check the label is actually populated after setup. Open Merchant Center, filter by the attribute, and confirm the count matches your expectation. A rule that silently matched nothing is a common and expensive failure.

Rebuild campaigns around the label once it is live. The label does nothing on its own. The value comes from the listing group structure and the bid strategy you build on top of it.

‍

Frequently asked questions

‍

How many custom labels can I create? Five per product, numbered 0 to 4. Each attribute supports up to 1,000 unique values, capped at 5,000 across the account.

Do custom labels affect which searches my Shopping ads appear for? No. They have no influence on query matching and are not visible to shoppers. Product title, description and attributes drive matching.

Can I set custom labels based on ROAS using feed rules? No. Feed rules can only reference data already present in your product feed. Google Ads performance data is not in the feed, so this requires an external tool.

How often should custom labels update? Attribute based labels such as season or product type can update on the normal feed schedule. Performance based labels should update daily, because a product's classification can change within a week.

What is a labelizer? A labelizer is a tool that reads Google Ads performance for every product, sorts each item into a segment, and writes that segment into a custom label field. The term comes from ProductHero's product name and is now used for the category as a whole. A labelizer differs from a feed manager because it reads campaign results rather than catalog attributes.

Do I need a feed management tool as well? They solve different problems. A feed manager handles getting clean product data into ad platforms. A bucketing tool handles keeping performance based labels current. Many accounts run both.

Can I use custom labels in Performance Max? Yes. Performance Max, Shopping and Demand Gen campaigns all support the attribute, which makes it one of the few levers for imposing structure on a PMax campaign.

‍

Where to start

Pick the simplest method that survives your catalogue.

If you have never used custom labels, start with a feed rule on price or product type. It costs nothing, takes twenty minutes, and proves the segmentation is worth building on before you add a tool.

If you already run feed rules and keep hitting the same wall, that your best segmentation ideas depend on performance data the feed does not contain, that is the point where automated bucketing earns its place. Not before.

The method matters less than the discipline. A single well maintained label used properly will improve campaign performance more than five labels nobody has looked at since setup.

‍