GMC, feed hygiene, and the Shopping moat

Why feed work is the cheapest 30% of ROAS

The feed work nobody wants to do is the work that compounds. Title structure, attribute completeness, custom labels, image policy, identifier accuracy. Boring. High-leverage. The brands that win Shopping in 2026 do this work weekly, not quarterly.

The reason this is undervalued: it doesn't feel like 'real' marketing. There's no creative, no bid strategy debate, no campaign architecture. Just a spreadsheet of attributes. But Smart Bidding consumes whatever is in the feed, and feeds with strong attributes outperform feeds with weak attributes by 20-30% on the same spend.

Title rewrite, the 5-attribute pattern

1. 1

   Lead with the buyer's search term

   Buyers search by category + attribute, not by brand name. Title format: [Material] [Type] [Brand] [Color] [Size]. Example: 'Stoneware Coffee Mug Set / Ad-Lab Co. / Cream / 12oz' instead of 'Ad-Lab Co. Mug Set'.

2. 2

   Front-load specificity

   First 60 characters of the title get the most weight. Pack them with the attributes Google's matching algorithm uses (material, style, dimension, intended use).

3. 3

   Avoid keyword stuffing

   Specific is good; spammy is a disapproval. Don't repeat the category twice. Don't list every imaginable use case. Google penalises stuffed titles.

4. 4

   Match titles to GTIN/MPN data

   If a product has a GTIN, use it. Mismatches between title and GTIN data are a top suspension trigger.

5. 5

   A/B test for top SKUs

   For your top 20 SKUs, test two title variants for 14 days. Run only one test at a time per SKU. The winner becomes the new baseline.

Feed attribute completeness checklist

100% completeness on the top half of this list is non-negotiable. The second half is where you separate from competitors.

Templates + glossary for feed work

Custom labels are the bidding control surface

Google Merchant Center exposes five custom_label fields (custom_label_0 through custom_label_4). They're free-text. You define what each one means. Smart Bidding uses them as splitting criteria when you tell it to.

Most accounts use one or two custom labels. The accounts that scale use all five. They become the lever that lets you split asset groups by margin, seasonality, velocity, lifecycle stage, and inventory position without rebuilding the campaign every time the catalog changes.

The five-label convention we run on every account

Reorder if it suits your business. Just don't leave any of them empty.

Wiring custom labels into asset groups

1. 1

   Map each SKU to all five labels in the feed

   Use feed rules in GMC or your feed-management tool. Each SKU gets a value for every label.

2. 2

   Build asset groups by margin × velocity

   High-margin × fast-velocity gets the most aggressive POAS target. Low-margin × slow-velocity gets the loosest. Mid-tier groups in between.

3. 3

   Use seasonality as a budget gate

   When a season starts, raise budget caps on that label's groups. When it ends, lower. Don't let evergreen-priority budgets stay on out-of-season SKUs.

4. 4

   Pause out-of-stock automatically

   Custom_label_4 = 'out' triggers a feed rule that excludes the SKU from all asset groups. Prevents bidding on inventory you can't ship.

Reference reading

Why most brands take 2-3 weeks to recover

GMC suspensions usually take brands 2-3 weeks to recover not because Google is slow, but because the brand spends the first 10 days arguing with the wrong policy. Suspension reasons are listed; you must respond to the exact reason cited, not your interpretation of why you got hit.

Brands that recover inside 48 hours follow a tight protocol: read the policy violation literally, fix the violation in the catalog or site, submit a single comprehensive reinstatement request that cites the exact policy section. No long emails, no general protests, no resubmits before the catalog is actually clean.

The 48-hour recovery protocol

1. 1

   Hour 0: read the policy violation

   Open GMC, click into the suspension, read the exact policy section cited. Write it down word-for-word. This is the only thing that matters.

2. 2

   Hour 1-4: audit the catalog against that exact policy

   Don't audit broadly. Audit specifically against the cited policy. If the policy is 'misrepresentation', look at every claim in product descriptions. If 'unclear pricing', look at price/sale_price/shipping consistency.

3. 3

   Hour 4-12: fix the violations

   Edit the catalog. Update PDP copy if needed. Push the corrected feed. Take screenshots of every fix.

4. 4

   Hour 12-24: file the reinstatement request

   Single request. Cite the exact policy section. List the specific fixes made (which SKUs, what changed). Attach screenshots. Tone is factual, not pleading.

5. 5

   Hour 24-48: monitor + respond once

   Google usually responds within 24-48 hours. If they reject, ask once for the specific remaining issue. Never argue about whether the original suspension was warranted.

The five most common suspension reasons

Templates + glossary for recovery