Merchant Center & Product Feed Health Checklist

Know which platform owns title, brand, GTIN, MPN, price, availability, taxonomy, variant data and channel-specific enrichment.

Incorrect values should have an accountable owner instead of being permanently patched downstream.

Offer IDs should not change casually; ID migrations need deliberate planning.

Price, stock and time-sensitive attributes need a refresh model that reflects catalogue volatility.

Teams should be able to connect a field/mapping change to downstream impact.

Do not omit known GTINs and do not invent them for products that genuinely have none.

Size, colour and other manufacturer variants should not incorrectly share one identifier.

Use the retailer name only where it genuinely manufactures/owns the product brand.

Do not fill MPN with arbitrary retailer SKUs for branded goods.

It is not a workaround for data that simply has not been collected.

A shopper and a platform should understand the item without decoding marketing language.

The most useful differentiators should survive truncation.

Different product families need different information priorities.

Use size, colour, capacity or other variant data where it helps identify the exact product.

Avoid keyword repetition and promotional filler.

Avoid placeholders, promotional overlays and wrong-variant imagery.

Use current taxonomy values and map deliberately.

It should support reporting, rules and segmentation rather than act as a catch-all.

Examples include colour, size, material, pattern, age group, capacity and dimensions.

Group related offers correctly while preserving accurate variant-level values.

Fix customer-facing catalogue structure where that is the real problem.

Check sale periods, tax treatment, membership logic and exact variant selection.

Customers should be able to buy what the listing says is available.

JSON-LD or microdata should not expose stale values.

Know the lag from source change to feed, submission, processing and page update.

Settings and product-level exceptions need to be commercially accurate.

They do not replace regular accurate product-data updates.

Teams know which source adds/removes products and which only enriches them.

Important transformations need an owner and a business reason.

Shared factual product data should normally be fixed upstream.

Preview representative products and edge cases.

Cascading transformations can create unexpected outcomes.

Prioritise blockers affecting important products and destinations.

Compare processed data with source, landing page and structured data.

They often require different owners and review processes.

Keep evidence of the change before asking Google to reassess.

Do not wait for campaign performance to reveal catalogue failure.

Measure recurrence and usable catalogue coverage over time.

Keep enough history to understand what changed.

Separate blockers, recurring operational debt, optimisation and architecture work.