Custom labels in your feed: which of the five fields are worth using

A label is only worth it if every value triggers a decision

Custom labels are five fields in the feed, custom_label_0 to custom_label_4, with which you filter products for Shopping, Performance Max and Demand Gen, for reports and for bids. Customers do not see them. There are no more than five, and each field holds exactly one value per product.

The rule: before a field is filled, it is settled for every possible value which campaign, which target ROAS or which budget follows from it. If that question cannot be answered, the label is information, not control, and the information is usually already in another attribute or in the shop.

Guides from feed tools such as DataFeedWatch or Lengow collect uses instead. Technically almost anything works, but commercially that is the wrong question, because five fields are scarce and every field you fill costs maintenance effort and, in Performance Max, campaigns.

Which four fields carry a decision

From a retailer's point of view, there are four attributes that regularly lead to a different setting in the account. The fifth field stays free.

The margin band is the strongest field, because it translates directly into the permissible ROAS. A product with a 20% contribution margin needs a different break-even from one with 45%. How to cut the bands and derive the target values is covered in Implementing target ROAS by margin group. It almost always earns a place, unless you already pass margin as the conversion value.

Stock and clearance trigger budget decisions: stop pushing goods with low stock, deliberately clear remaining stock with a lower target, exclude sold-out lines. More in Steering Google Ads by stock position.

Season is worth it if you deliberately ramp seasonal goods up or down at certain times. A shop with the same range all year round does not need this field.

I usually regard the price band as a reporting field. It shows whether cheap products tie up clicks without bringing contribution margin. It only justifies its own campaign if it leads to a different target ROAS, and then it is often really a margin or basket topic.

The reserve is not a luxury. If you permanently assign all five fields, every promotion means reworking an existing label and moving products between campaigns in the process.

What a label really controls in Performance Max

In Performance Max, according to Google, the AI sets the bids based on the goal of the campaign. Listing groups only determine which products run in an asset group, and they have no bid of their own. A label in a listing group therefore changes nothing about how high the bid is for those products.

A different target ROAS per label value is only possible through separate campaigns. There is no portfolio bid strategy for Performance Max, and target ROAS is tied to the individual campaign. The same goes for budget.

At the same time, Google expressly recommends grouping products via custom labels rather than creating very many listing groups. Per asset group, up to 1,000 listing groups are possible, but not recommended. In PMax, labels are therefore the tool for assigning products cleanly to a campaign. They are no good for bids within a campaign. How to cut asset groups within a campaign is a different question, see Cutting asset groups sensibly.

It is different in standard Shopping. There, every product group has its own bid, and all products in it bid the same. A label can carry a bid decision there without a new campaign, as long as you work with manual CPC bids per product group.

The calculation: how many campaigns a label matrix costs

Every label value that needs its own target costs a campaign in Performance Max. If you combine fields, it multiplies.

Worked example with round, invented figures: suppose a shop achieves 300 conversions a month in PMax. Three margin bands give three campaigns with 100 conversions on average. If it combines the margin bands with two stock statuses, that is six campaigns with 50 conversions on average, and because revenue is never evenly distributed, individual campaigns sit well below that. Add two season values and it is twelve campaigns with 25 on average.

The rule I derive from this: only one field determines the campaign structure, usually the margin band. Stock and season work as an exclusion or as a temporary move into an existing campaign, not as a dimension of their own. How much data a campaign needs for a stable target is covered in How many conversions target ROAS needs.

Maintaining labels automatically without losing control

Maintaining labels by hand in Merchant Center does not hold up in any range with changing goods. Google recommends feed rules to set labels from values already in the feed, such as price ranges. For attributes that are not in the feed, such as margin or stock cover, you need an additional source. A Google Sheets source syncs every 24 hours by default, and the schedule can be changed. The margin itself does not have to go into the feed, only the band.

You need to allow for three delays:

• According to Google, new or changed labels take up to 24 to 48 hours to appear in Google Ads.
• In standard Shopping, Google does not automatically carry changed label values over into the product groups. A new value has to be added to the structure, otherwise the products end up in "Everything else".
• If a product changes its label, in PMax it changes campaign, with a different target and a different budget.

That is why I advise thresholds with a gap. Suppose a product becomes "low" at under 14 days of stock cover and only goes back to "normal" at over 30 days. Without that gap, products at the boundary flip between two campaigns every day.

Which labels you can leave out

Per label, up to 1,000 unique values are allowed in the account, 5,000 in total, each value 1 to 100 characters. Values beyond that are not taken into account for reports and bids, and the products count as unlabelled. Labels with product names, item IDs or fine-grained category paths quickly hit this limit in large ranges and are superfluous anyway, because separate attributes exist for them.

I would also leave out labels that only describe: colour, material, target group. As long as nobody derives a different campaign or target from them, they are usually just as visible in reports through other attributes.

Performance labels such as "bestseller" or "zombie" are a special case. They trigger a decision, but they measure the result of the bids they are meant to steer. When that helps and when it reinforces itself is covered in Segmenting products by performance.

By the way, a product does not have to have a value in every label. An empty season field for year-round goods is not an error.

When the rule does not apply

A small shop with a single PMax campaign can use labels purely for reports, with no consequence for structure. That is fine as long as it is clear that they change nothing about bids there. Anyone running mostly standard Shopping with product group bids can use more fields as a bid dimension, because no new campaign is needed. And anyone who already passes margin as the conversion value often no longer needs the margin band for target ROAS, because the bid strategy then sees the differences in the value itself. The field then becomes free for stock or season.

As of Google features: November 2026.