Shopping feed optimisation: the seven fields that move ROAS
Most product feeds leak revenue in the same seven places.
Feed work is unglamorous and high-leverage. Sharpening seven fields is usually worth a 15–25% ROAS lift before you touch a bid. The brands that ignore feed quality and pour budget into smart bidding are paying Google to optimise around bad inputs — and getting exactly what they asked for.
The seven fields
- title — front-load brand, then primary attribute (e.g. 'Nike Air Zoom Pegasus 41 Men's Running Shoe').
- GTIN — accurate and complete; missing GTINs cap impression share on competitive products.
- product_type — your taxonomy, not Google's, used for bid segmentation.
- google_product_category — Google's taxonomy, kept at the deepest accurate level.
- custom_label_0 — margin tier (high/mid/low).
- custom_label_1 — performance tier from last 30 days (hero/scaling/long tail).
- custom_label_2 — inventory status or stock weeks remaining.
The cadence
Quarterly feed audit, monthly title experiments via a feed management tool like Feedonomics or Channable. Measure on category-level ROAS, not blended — a feed change that lifts hero ROAS by 8% and tanks long-tail ROAS by 30% will look flat in aggregate.
What to skip
Don't waste cycles on description (Shopping barely reads it), on color when there's no shopper intent for it, or on adding every possible product attribute. Marginal completeness beyond the seven above rarely moves the needle.
