How to Use Merchant Center Custom Labels

Start with a written label dictionary instead of filling five fields because they exist. Assign one business dimension to each field, such as season in custom_label_0 and margin band in custom_label_1. Do not mix season, promotion status, and lifecycle stage inside the same field. Write the control before changing production: the label dictionary, allowed values, owner, and refresh rule. Name the accountable owner and preserve the decision date so later reporting uses the rule that actually existed.

Google permits custom_label_0 through custom_label_4 and one value per product for each label. Keep values short, consistent, and operational. A value such as high_margin needs an internal definition, owner, and refresh cadence before it can support a bidding or reporting decision. Retain sample product records and the approved dictionary as the evidence set. Verify exceptions separately, and do not infer ranking, demand, savings, or causal business impact from the configuration or report alone.

A useful label lets a team group products for a real campaign action or report question. Season, clearance state, selling rate, launch cohort, and price band can be useful when the definitions are explicit. Avoid labels that merely repeat product type, brand, or another dependable feed attribute. Write the control before changing production: the decision the label is allowed to influence. Name the accountable owner and preserve the decision date so later reporting uses the rule that actually existed.

Write the decision beside the proposed field: which products will be separated, what review will occur, and when the label expires. If no owner would bid, budget, report, or investigate differently because of the value, leave the optional field empty rather than creating metadata debt. Retain a before-and-after inventory of affected item IDs as the evidence set. Verify exceptions separately, and do not infer ranking, demand, savings, or causal business impact from the configuration or report alone.

Add the chosen value through the primary data source, a supplemental source, the Merchant API, or a governed rule. Google documents format differences across text, XML, and API sources, so test the exact ingestion path used by the catalog rather than copying an example blindly. Write the control before changing production: one authoritative writer for every custom-label field. Name the accountable owner and preserve the decision date so later reporting uses the rule that actually existed.

Use feed rules only when the source attribute is reliable and the mapping is deterministic. Normalize capitalization and whitespace before upload. Prevent two systems from competing to populate the same custom-label field, and preserve the upstream value that produced each submitted result. Retain the submitted row, processing result, and source attribute as the evidence set. Verify exceptions separately, and do not infer ranking, demand, savings, or causal business impact from the configuration or report alone.

Confirm that representative products show the intended values after processing and that disapproved or excluded items are not mistaken for a label failure. Check blank values, unexpected variants, stale cohorts, and products that moved between business states before building campaign subdivisions. Write the control before changing production: validation before Google Ads structure changes. Name the accountable owner and preserve the decision date so later reporting uses the rule that actually existed.

Google sets limits on the number and length of custom-label values. Keep the vocabulary bounded and watch for near-duplicates such as high-margin, high_margin, and High Margin. A clean value list makes campaign grouping reproducible and reduces silent fragmentation. Retain processed-item views and a distinct-value export as the evidence set. Verify exceptions separately, and do not infer ranking, demand, savings, or causal business impact from the configuration or report alone.

After the values sync, use them to create product subdivisions in Shopping or feed-based Performance Max campaigns. Apply a complete partition so products without the chosen value remain visible in an explicit remainder group instead of disappearing from the operator's mental model. Write the control before changing production: a complete product partition with an explicit remainder. Name the accountable owner and preserve the decision date so later reporting uses the rule that actually existed.

Name campaign groups with the field and value, not an interpretation such as winners. Labels organize inventory; they do not prove profitability or demand. Keep bids and budgets unchanged during the first sync check unless the planned experiment specifically requires another controlled change. Retain Google Ads product-group membership and item counts as the evidence set. Verify exceptions separately, and do not infer ranking, demand, savings, or causal business impact from the configuration or report alone.

Schedule a recurring comparison between the catalog, Merchant Center, and Google Ads grouping. Investigate products that are unlabeled, unexpectedly relabeled, duplicated across intended cohorts, or retained after a promotion ends. Separate feed latency from a true rule defect before editing campaign structure. Write the control before changing production: drift thresholds and an escalation owner. Name the accountable owner and preserve the decision date so later reporting uses the rule that actually existed.

When a definition changes, version the rule and record the effective time. Do not recycle an old value for a new meaning because historical reports will become ambiguous. Retire a value deliberately and confirm that dependent campaign groups no longer rely on it. Retain dated exports from catalog, Merchant Center, and Ads as the evidence set. Verify exceptions separately, and do not infer ranking, demand, savings, or causal business impact from the configuration or report alone.

Compare like-for-like dates, products, eligibility, inventory, prices, and campaign settings. A label can expose a useful segment, but a better result for that segment does not prove the label or the labeled characteristic caused the difference. Use controlled experiments for causal questions. Write the control before changing production: the comparison window and known confounders. Name the accountable owner and preserve the decision date so later reporting uses the rule that actually existed.

Keep cost, conversion, margin, returns, and stock signals at their correct scopes. Annotate launches, promotions, outages, and feed changes. The operating outcome is a reliable product grouping that supports decisions, not a claim that a naming convention produced incremental revenue. Retain segment definitions, annotations, and reconciled reports as the evidence set. Verify exceptions separately, and do not infer ranking, demand, savings, or causal business impact from the configuration or report alone.