Google Merchant Center Framework

Google Merchant Center analysis and optimization framework. Use when the user asks about Shopping feed health, product disapprovals, title optimization, pric...

Security Scan

This is an instruction-only framework for diagnosing and optimizing Google Merchant Center feeds; its requirements and instructions are consistent with that purpose and it does not request unexpected credentials or install code.

Purpose & Capability

The name/description (GMC analysis & optimization) matches the SKILL.md content: questions to collect context, diagnostic frameworks, scoring, and optimization guidance. There are no unrelated requested binaries, env vars, or config paths.

Instruction Scope

Runtime instructions are limited to collecting user-provided context, asking follow-ups, applying diagnostic frameworks, and recommending fixes. The doc notes optional live access via a Google Merchant Center connection but does not instruct the agent to read local files, environment variables, or send data to unexpected endpoints.

Install Mechanism

No install spec and no code files — this is instruction-only, so nothing is downloaded, written to disk, or executed by the skill itself.

Credentials

The skill declares no required environment variables, no primary credential, and no config paths. The Integration Note reasonably references using a Merchant Center connection for live data; that implies credentials would be supplied by whatever MCP integration the platform provides, but the skill itself does not request unrelated secrets.

Persistence & Privilege

always is false and there is no install behavior that modifies agent config or requests persistent presence. Normal autonomous invocation is allowed by platform defaults but the skill does not escalate privileges.

Assessment

This skill is an offline decision framework for Google Merchant Center and appears coherent. Before using it with live data, connect to your Merchant Center only via the platform's official OAuth/connector (do not paste long account credentials into chat). Prefer least-privilege access scopes, and avoid pasting full feeds or customer PII in chat; instead provide redacted exports or allow the platform connector to fetch data. If you intend the agent to act on your GMC account, verify which scopes/actions it will be granted and revoke access when finished.

Diagnose feed health, optimize product listings, resolve disapprovals, analyze price competitiveness, and assess Shopping campaign readiness. A decision-framework for agents working with Google Merchant Center accounts of any size.

How This Skill Works

Step 1: Collect context from the user's message — merchant category, product count, known issues, goals (fix disapprovals, improve performance, launch Shopping ads, etc.).

Step 2: Ask one follow-up with all remaining questions using multiple-choice format. Key questions:

• a) Account size: (1) <500 SKUs, (2) 500-10K, (3) 10K-100K, (4) 100K+
• b) Primary goal: (1) Fix disapprovals, (2) Improve feed quality, (3) Boost Shopping performance, (4) Launch new campaign, (5) General audit
• c) Feed method: (1) Direct API, (2) Scheduled fetch (XML/CSV), (3) Shopify/WooCommerce plugin, (4) Supplemental feeds, (5) Not sure
• d) Current disapproval rate: (1) <2%, (2) 2-10%, (3) 10-25%, (4) >25%, (5) Unknown

Allow shorthand answers (e.g., "1a 2c 3b 4e").

Step 3: Analyze using the frameworks below. Prioritize by impact: disapprovals first (products not showing), then data quality (products showing poorly), then performance (products showing but underperforming).

Step 4: Deliver structured output with specific fixes, benchmarks, and priority order.

Integration Note

This skill works best when paired with a Google Merchant Center MCP for live data access. Without live data, the agent applies frameworks to data the user provides (exports, screenshots, descriptions of issues). With live data, the agent can pull product statuses, diagnostics, and price insights directly.

---

1. Feed Health Analysis

Health Score Framework

Calculate a feed health score from these weighted components:

Component	Weight	How to Score
Approval rate	30%	100% approved = 100 pts. Deduct 2 pts per % disapproved
Required attribute completeness	25%	% of products with all required attrs filled
Recommended attribute completeness	15%	% of products with 8+ of 12 recommended attrs
Data freshness	15%	Updated <24h = 100, 24-48h = 70, 48-72h = 40, >72h = 0
Image quality	15%	% of products with compliant images (no watermarks, no placeholders, >100x100px)

Score interpretation:

• 90-100: Excellent. Focus on performance optimization.
• 75-89: Good. Fix remaining disapprovals and fill attribute gaps.
• 50-74: Needs work. Systematic issues present. Prioritize disapprovals, then required attributes.
• <50: Critical. Feed is substantially broken. Triage disapprovals by volume.

Required vs Recommended Attributes

Always required (all products):

• id, title, description, link, image_link, availability, price

Required for most categories:

• brand (required except for movies, books, musical recordings)
• gtin (required for all products with a manufacturer-assigned GTIN)
• condition (required if used or refurbished)

Conditionally required:

• color, size, gender, age_group — required for Apparel & Accessories
• item_group_id — required for product variants
• shipping — required if not set at account level
• tax — required in the US if not set at account level

High-impact recommended attributes:

• additional_image_link (up to 10 images; products with 3+ images get ~15% higher CTR)
• product_type (your own categorization; helps with campaign structure)
• sale_price + sale_price_effective_date (triggers sale badge in Shopping)
• custom_label_0 through custom_label_4 (essential for campaign segmentation)
• product_highlight (up to 10 bullet points; shown in free listings)
• product_detail (section_name + attribute_name + attribute_value tuples)

Diagnostic Priority Order

1. Account-level suspensions — nothing shows until resolved
2. Item-level disapprovals — products removed from serving
3. Item-level warnings — products serve but with reduced visibility
4. Missing required attributes — will become disapprovals
5. Missing recommended attributes — reduces competitiveness
6. Data quality issues — poor titles, descriptions, categorization

---

2. Product Optimization

Title Optimization

Titles are the single highest-impact attribute for Shopping performance. Google uses titles for query matching more heavily than descriptions.

Structure formula by category:

Category	Title Formula	Example
Apparel	Brand + Gender + Product Type + Attributes (Color, Size, Material)	"Nike Women's Air Max 270 Running Shoes - Black/White, Size 8"
Electronics	Brand + Product Line + Model + Key Specs	"Samsung Galaxy S24 Ultra 256GB Titanium Black Unlocked"
Home & Garden	Brand + Product Type + Material + Key Dimensions + Color	"KitchenAid Classic 4.5-Quart Tilt-Head Stand Mixer Onyx Black"
Beauty	Brand + Product Line + Product Type + Size/Count + Variant	"CeraVe Moisturizing Cream Face and Body 19oz Fragrance-Free"
Grocery	Brand + Product Name + Flavor/Variant + Size/Count + Pack	"KIND Bars Dark Chocolate Nuts & Sea Salt 12-Count Box"

Title rules:

• Max 150 characters; first 70 characters are most critical (truncation in mobile)
• Front-load the most important keywords in the first 70 characters
• Never use ALL CAPS or promotional text ("Free Shipping", "Best Price")
• Include color, size, material when relevant — these are matching signals
• Use the actual brand name, not abbreviations
• Separate attributes with hyphens or commas, not pipes

Title quality scoring:

Signal	Points
Brand name present and correct	+20
Product type keyword included	+20
Key differentiating attribute (color/size/material)	+15 each, max +30
Within 70-character sweet spot	+15
No promotional language	+15
Total possible	100

Description Optimization

• 1,000-5,000 characters optimal (minimum 500)
• First 150-200 characters most critical (may be shown in free listings)
• Include keywords naturally — Google does use descriptions for matching, but less than titles
• Include specifications, materials, dimensions, use cases
• No HTML tags (stripped by Google), no promotional language, no links

Image Requirements

Requirement	Standard	Notes
Minimum resolution	100x100px (250x250 for apparel)	800x800+ recommended for zoom
Max file size	16MB
Format	JPEG, PNG, GIF, BMP, TIFF, WebP
Background	White or transparent preferred	Non-white OK but clean background required
Watermarks	Not allowed	Automatic disapproval
Text overlay	Not allowed on main image	OK on additional images
Promotional overlay	Not allowed	"Sale", "Free shipping" overlays = disapproval
Product visibility	Product must fill 75-90% of frame	No tiny product in large frame

Image optimization checklist:

• Main image: product only, white background, high resolution
• 3+ additional images (lifestyle, different angles, scale/size reference)
• Apparel: show product on a person or flat-lay (not hanger)
• Consistent image style across product line

GTIN/MPN Completeness

• GTIN is critical. Products with valid GTINs get ~20% more impressions than those without.
• Google cross-references GTINs against the GS1 database. Invalid GTINs cause disapprovals.
• If a product genuinely has no GTIN (custom, handmade, vintage, parts), set identifier_exists to no.
• Never fabricate GTINs. Never reuse GTINs across different products.
• MPN is required when GTIN is not available and the product has a manufacturer part number.

---

3. Price Competitiveness

Interpreting Price Benchmarks

Google provides price benchmarks in the Price Competitiveness report. Key metrics:

Metric	What It Means	Action Threshold
benchmark_price	Median price of the same product from other merchants	If your price >15% above: expect significantly lower CTR
price_difference_percentage	Your price vs benchmark	>10% above: review pricing. >20% above: likely suppressed
country_code	Market the benchmark applies to	Compare only within same market

Competitive Visibility Score Framework

Position	Your Price vs Benchmark	Expected Impact
Strong	>10% below benchmark	High impressions, high CTR. Verify margin is acceptable.
Competitive	Within +/-10% of benchmark	Normal serving. Optimize other signals.
Weak	10-20% above benchmark	Reduced impressions. Consider sale_price or promotions.
Uncompetitive	>20% above benchmark	Severely reduced serving. Re-evaluate pricing or exclude from Shopping.

Price Strategy Decisions

When price is too high vs benchmark:

1. Can you lower the price? Do it.
2. Can you add a promotion or sale_price? Triggers sale badge, improves CTR even if final price is still above benchmark.
3. Can you differentiate on value (bundle, warranty, fast shipping)? Add to title and description.
4. Is the product a loss leader for competitors? Consider excluding from Shopping.
5. Use custom_label to segment high-priced items into separate campaigns with different ROAS targets.

Price-Landing Page Match — #1 cause of avoidable disapprovals:

• Price shown on page must match price attribute exactly (including currency)
• sale_price must match the visible sale price on the landing page
• If microdata/structured data on the page differs from the feed, the feed value may be overridden or flagged

---

4. Performance Analysis

Shopping Performance Benchmarks by Category

Category	Median CTR	Good CTR	Median CPC (Shopping Ads)	Median Conv. Rate
Apparel & Accessories	1.2%	>2.0%	$0.40-0.70	1.5-2.5%
Electronics	0.8%	>1.5%	$0.50-1.00	1.0-2.0%
Home & Garden	1.0%	>1.8%	$0.35-0.65	1.5-3.0%
Health & Beauty	1.5%	>2.5%	$0.30-0.55	2.0-3.5%
Grocery & Food	1.8%	>3.0%	$0.20-0.40	2.5-4.0%
Toys & Games	1.3%	>2.2%	$0.25-0.50	2.0-3.5%
Sports & Outdoors	1.0%	>1.7%	$0.35-0.65	1.5-2.5%
Auto Parts	0.7%	>1.2%	$0.40-0.80	1.0-2.0%

Performance Diagnostic Tree

Low impressions:

1. Check disapprovals — products not serving at all?
2. Check price competitiveness — priced out of auctions?
3. Check title relevance — titles match search queries?
4. Check bid/budget (paid) — budget exhausting before end of day?
5. Check product ratings — competitors with ratings outranking you?

Low CTR (impressions OK):

1. Image quality — compelling vs generic?
2. Title quality — informative vs vague?
3. Price position — higher than competitors in same carousel?
4. Sale badge — competitors showing sales, you are not?
5. Shipping speed/cost — competitors showing "Free delivery" and you are not?

Low conversion rate (clicks OK):

1. Landing page match — does the page show what the listing promised?
2. Landing page speed — >3 seconds load time loses ~50% of mobile users
3. Price on landing page — any surprise fees or shipping costs?
4. Mobile experience — is the landing page mobile-optimized?
5. Availability — "Add to cart" immediately visible?

Key Metrics to Track Weekly

Metric	Red Flag
Disapproval rate	>5% or increasing trend
Impression share (paid)	<50% in target categories
CTR	Below category median
Click-to-conversion rate	<1% for most categories
Price competitiveness %	>30% of products uncompetitive

---

5. Disapproval Workflows

Top 10 Disapproval Reasons and Fix Paths

#	Reason	Cause	Fix	Timeline
1	Misrepresentation	Misleading claims, missing business info	Remove superlatives, add return policy, add About Us page	3-7 days, may need manual review
2	Price Mismatch	Feed price ≠ landing page price	Sync feed price to page exactly; check currency	24-72h re-crawl
3	Availability Mismatch	Feed says in stock, page says out of stock	Sync inventory feed; increase update frequency	Next crawl
4	Missing GTIN	Has manufacturer GTIN but not in feed	Add GTIN or set identifier_exists: no	Immediate on next processing
5	Image Overlay	Watermarks, promo text on main image	Replace with clean product image	24-72h re-crawl
6	Landing Page Error	404/500 errors, geo-blocking Googlebot	Fix URLs; unblock Googlebot	Next crawl
7	Insufficient Product Data	Title/description too short or generic	Title >30 chars with brand+type; description >500 chars	Immediate
8	Missing Shipping Weight	Carrier-calculated rates but no weight in feed	Add shipping_weight or switch to flat-rate	Immediate
9	Restricted Product	Regulated category without verification	Verify compliance; apply for merchant verification if required	Varies
10	Duplicate Item ID	Same ID reused for different products	Assign unique, stable IDs per variant	Immediate

Disapproval Triage Matrix

Volume	Same Reason?	Action
>100 products	Yes	Systematic root cause — fix feed template or feed rule, not individual items
>100 products	No	Group by reason, fix highest-volume first
10-100 products	—	Fix in bulk via supplemental feed or feed rules
<10 products	—	Fix individually; watch for pattern

Appeal Process

1. Fix the issue first. Never appeal without fixing.
2. Wait 24-72 hours for automatic re-crawl.
3. If not resolved: Request manual review (Diagnostics > Item Issues > Request Review).
4. One appeal at a time. 7-day cooldown applies for repeated policy violations.

---

6. Feed Management

Feed Method Decision Tree

Platform has native GMC integration (Shopify, WooCommerce)?
  YES → Use platform plugin as primary feed
        Need to override attributes the plugin doesn't support?
          YES → Add supplemental feed for overrides
          NO  → Plugin-only is sufficient
  NO  → Have developer resources?
        YES → Content API (real-time; best for >10K SKUs or fast inventory)
        NO  → Scheduled fetch (XML/CSV hosted on your server; update at least daily)

Feed Types

Feed Type	Use Case	Update Frequency
Primary feed	All products, all required attributes	Daily minimum; every 6h for fast-moving inventory
Supplemental feed	Override/add attributes without touching primary	As needed
Feed rules	Transform data at processing time (regex, lookups)	Applied every processing run
Content API	Real-time individual product updates	Real-time; 100K requests/day limit
Automated feeds	Google crawls your site	Google's schedule (not your control)

Supplemental Feed Strategy

Use supplemental feeds for:

• Custom labels — campaign segmentation (margin tiers, seasonality, priority)
• Title A/B testing — optimized titles without touching your CMS
• Missing GTINs — lookup table mapped to product IDs
• Promotion IDs — linking products to active promotions
• Seasonal overrides — holiday titles, temporary attribute changes

Supplemental feeds match on id — must exactly match the primary feed id.

---

7. Shopping Campaign Readiness

Pre-Launch Checklist

Check	Requirement
Feed health score	>80 (Section 1 framework)
Disapproval rate	<5% of active products
Price competitiveness	>60% of products within +10% of benchmark
Landing page speed	<3 seconds, mobile-friendly
Conversion tracking	Google Ads tag firing, revenue tracking verified
Shipping info	Accurate in feed or account settings

Standard Shopping vs Performance Max

Factor	Standard Shopping	Performance Max
Query control	High (negatives, search term reports)	Low (limited insights)
Bidding	Manual CPC or Target ROAS	Automated only
Best for	Tight ROAS targets, query-level strategy	Broad reach, new advertisers, >30 conv/month
Min daily budget	$5-10/day viable	$50-100/day recommended
Ramp-up	Immediate	2-4 week learning phase

Custom Label Strategy

Slot	Use	Example Values
custom_label_0	Margin tier	high_margin, medium_margin, low_margin
custom_label_1	Product priority	hero, standard, long_tail
custom_label_2	Seasonality	evergreen, summer, holiday, clearance
custom_label_3	Price range	under_25, 25_to_100, over_100
custom_label_4	Performance tier	top_seller, average, underperformer, new

ROAS Targets by Category (Starting Points)

Category	Conservative ROAS	Aggressive ROAS
Apparel	400-600%	200-300%
Electronics	600-800%	300-500%
Home & Garden	400-600%	200-400%
Health & Beauty	300-500%	150-300%
Grocery	200-400%	100-200%

---

Output Format

• Start with a Feed Health Score (0-100) and top-line summary
• Group findings into: Critical (disapprovals, policy) → High (data quality, price) → Medium (optimization) → Low (nice-to-have)
• Include specific product counts and percentages per issue
• Provide exact fix instructions, not general advice
• Mark estimates with ⚠️ when based on incomplete data
• End with a prioritized action plan (numbered, with effort: 5min / 30min / 2hrs / half-day)
• Include a 7-day and 30-day check-in plan