custom_label_0

Custom labels are the lever between "all products" and "Advantage+ should bid differently on these specific products". Without them, your ad campaigns target the whole catalog or hand-curated product lists. With them, your campaigns target dynamic segments driven by feed data — Champions (high-ROAS items), Cash Cows (high-margin items), New Arrivals, Promo, Stockout-Prone — and the segments update themselves as the underlying data changes.

The convention that works: each label slot encodes one stable dimension. `custom_label_0` for ROAS tier (Champion, Climbing, Stable, Underperforming). `custom_label_1` for margin tier (premium, standard, clearance). `custom_label_2` for inventory state (in-stock, low-stock, out-of-stock). `custom_label_3` for campaign-tagged groupings (BFCM-2026, summer-2026). `custom_label_4` for ad-hoc experiments. Pick a convention, document it, and don't deviate — once 50,000 SKUs are tagged, changing the scheme is painful because every campaign rule that references the slot has to be updated in lockstep.

The ROAS-tier pattern is the most common starting point and the most operationally valuable. Pull Google Ads metrics over a 30-day lookback window, calculate ROAS per SKU, bucket into four tiers based on category-specific thresholds: Champion (>4x ROAS), Climbing (2-4x), Stable (1-2x), Underperforming (<1x). The campaign then has four product groups with very different bid strategies. Champions get aggressive bids and broad audience exposure; Underperformers get conservative bids or get paused entirely. The same revenue gets earned with substantially less spend.

For Champion-tier items, the strategic question is whether to push more (gain more impressions and convert better) or maintain (preserve margin on already-converting items). Most accounts under-spend on Champions because manual bid management can't keep pace with daily ROAS shifts. The label-driven approach lets Smart Bidding pursue volume on Champions while protecting margin elsewhere.

Updates need to be automated. ROAS tiers should rebucket weekly from Google Ads metrics; margin tiers from your finance data; inventory states from your stock system. Manually re-tagging custom labels across a sizeable catalog is unsustainable, and the labels become stale within weeks. Tools that pull metrics, apply rules, and write labels back to the feed (Performance Labels does this) are how this scales. The cadence matters: weekly rebucketing catches genuine performance shifts; daily rebucketing chases noise.

For consistency across SKUs, the case-sensitivity rule matters more than most teams realise. "Champion" and "champion" and "CHAMPION" are three different values to Google's classifier. A product set rule "custom_label_0 = Champion" misses items tagged "champion". Pick one canonical casing — convention is Title Case — and enforce at the label-writing layer. Don't trust SKU-by-SKU consistency; enforce it programmatically.

Below the 30-SKU threshold per product group, Advantage+ and Smart Bidding don't have enough signal to optimise effectively. If your Champion tier has 20 items, Smart Bidding can't distinguish their performance from noise. This means category-specific thresholds for ROAS tier definitions — small high-volume categories use higher Champion thresholds (more selective) so the tier has fewer but more clearly-winning items; large low-velocity categories use lower thresholds (more inclusive) so the tier reaches scale.