Discussion starter: an entity SEO question, posted with an answer from Niraj Raut to open the thread. If you have dealt with this on a site, reply with what you saw, especially where it differs.
- Entity SEO
- Structured Data
How many Product markup warnings do you tolerate at scale, given many recommended fields don’t change eligibility?
Disclosure: Niraj Raut, who posted this answer, runs the SEO consultancy linked at the end of it.
1 reply
-
adminAdmin
I would not set the tolerance as a number. At scale, the warning count in Search Console says more about how many URLs share a template than about how much is wrong: one unmapped field on one template can produce hundreds of thousands of warnings, and one fix clears them all. The count is not even complete, because Google says the rich result reports show a sample of detected items. What I would set instead is a tolerance per class of issue: zero for anything that makes items invalid or makes the markup disagree with the page or the feed, a fix for recommended fields where you already hold the data, and documented acceptance for data you do not have or supply elsewhere.
The premise of the question holds, with one exception. A missing recommended property does not make an item ineligible. But some recommended properties can suppress a listing when they are present with a bad value, so “it’s only a warning” is true for missing fields and can be false for wrong ones.
What a warning does and does not mean
Search Console’s rich result reports (for products, “Merchant listings” and “Product snippets”) sort items into Valid and Invalid. An item with at least one critical issue is Invalid and cannot appear as a rich result. Non-critical issues (the warnings in the question) leave the item Valid. The same help page says the reports are not a comprehensive list of detected items but a sample, and that the counts are items rather than pages.
Google’s merchant listing documentation keeps the required set short:
name,imageand anOfferwithpriceandpriceCurrency, and merchant listing experiences need a price greater than zero. Everything else, includinggtin,brand,availability,itemCondition,shippingDetails,hasMerchantReturnPolicy,aggregateRatingandreview, is recommended, and Google frames recommended properties as ones that could provide a better user experience. Two details on that page change how warnings should be read:- It recommends a business-wide return policy and shipping policy under
Organizationmarkup, with the product-level properties used only for products whose policies differ. - It warns that a listing may not display if
priceValidUntilindicates a past date. A recommended field with a bad value can cost you the listing; a missing one does not.
Tolerance by class, not by count
Class Examples Tolerance Reason Critical issues on revenue templates Missing Offer or price, a price of zero, missing priceCurrencyNone The item is Invalid, and every Offer needs a price for merchant listings Markup that disagrees with the page or feed JSON-LD price, currency or availability that differs from what the page shows None Google says it may try to verify merchant listing data before showing it, and if you run a feed the disagreement can also surface as Merchant Center mismatch issues Bad values in recommended fields A past priceValidUntil, malformed datesFix in the current sprint Some can stop the listing from displaying Missing fields you hold gtin,brand,itemConditionon refurbished stock,descriptionFix at the mapping, once per template Cheap, and it gives Google identifiers and facts the page already carries Missing fields supplied elsewhere shippingDetailsandhasMerchantReturnPolicywhen policies sit in Organization markup, Merchant Center, or the Search Console shipping and returns settings added in November 2025Accept and document Google recommends a business-wide policy, with product-level markup for exceptions Missing fields you do not have aggregateRatingorreviewon products with no reviews,gtinon made-to-order productsAccept Filling the field now would mean inventing data Watch for this:
The last row is where teams chasing a zero count do damage. Placeholder GTINs, a default star rating or a boilerplate return policy that does not match your real terms all clear warnings by adding data you cannot stand behind. Invented identifiers can attach a product to the wrong catalogue entry, and markup that misrepresents the page can draw a structured data manual action. A cosmetic warning is a much smaller problem than either.
Turn the warning count into a template matrix
At scale, each warning belongs to an issue type and a template, and each of those pairs has one owner and one fix. Map them before deciding anything. Think of it as a leaking roof: counting drips on the floor tells you it is raining, while finding the three holes in the roof tells you what to repair.
Hypothetical example:
A retailer has 900,000 product URLs on three templates: standard products, marketplace-seller products and refurbished stock. The Merchant listings report shows five issue types across a large share of sampled items. Mapped to templates, the work is much smaller:
- Shipping and returns warnings on every template, while both policies are already configured at account level. Accept and document.
- Rating and review warnings on most products, because most have no reviews yet. Accept.
- Missing
gtinon the seller template only, because sellers are not asked for one. Fix through seller onboarding and accept the remainder. - Missing
itemConditionon the refurbished template. A one-line fix, worth doing because a refurbished item without a condition tells searchers less than the page does. - Past
priceValidUntildates on the promotions template. Fix now, because those listings may not display.
Hundreds of thousands of warnings become three template changes and two documented exceptions.
Monitor rates and changes, not the total
Because the report is a sample and its counts move with crawling, the total makes a poor alarm. These signals are more useful:
- For each template, the share of sampled items carrying each issue type, charted weekly, with an alert on step changes after deployments.
- Your own structured data crawl of a fixed, stratified sample (a few hundred URLs per template), comparing JSON-LD price, currency and availability against the visible page and, if you run one, the feed.
- Spot checks in URL Inspection or the Rich Results Test for templates that inject JSON-LD with JavaScript, to confirm what Google actually rendered.
- Invalid item counts in the Merchant listings and Product snippets reports, with a threshold of zero on revenue templates.
A sudden rise in a warning you had accepted still deserves a look. Accepted means “known and understood”, not “ignored”: a jump in missing
gtinon a template that used to be clean is a reason to check the PIM or feed mapping before assuming anything changed at Google.Does fixing recommended fields move clicks?
Public evidence is thin. Google frames recommended properties around user experience and display rather than ranking, and no controlled public test exists that I can point to that ties a warning class to rankings. The plausible path is appearance: shipping and returns details, genuine ratings and a clear condition can change how a result looks, and that can change click-through. That makes it testable, within limits.
- Hypothesis. Adding a field you hold (say
gtinoritemCondition) to part of a template changes how its results display and lifts CTR at comparable positions. - Groups. Treat half the categories on one template and hold the rest back, matched on traffic and seasonality.
- Metrics. CTR at comparable positions from Search Console, the share of treated items reported Valid, and, if you run a feed, any change in Merchant Center item issues.
- Duration. The change only counts once Google has recrawled the page. On a large catalogue the long tail can take weeks or months to be recrawled, so start with categories your logs show Googlebot visiting often, and read the result once most treated URLs have been recrawled rather than on a fixed date.
- Decision. If CTR moves with display changes, prioritise the fields that change what searchers see. If nothing moves, keep the markup minimal and accurate, and spend the engineering time elsewhere.
When the tolerance should be stricter or looser
- Stricter without a Merchant Center feed. If page markup is the main structured source Google has for your products, the recommended fields you hold carry more of the load.
- Stricter on marketplaces. Completeness varies by seller, so set thresholds per seller cohort and make identifiers part of onboarding rather than chasing them listing by listing.
- Looser for feed-first retailers whose shipping, returns and identifiers are maintained in Merchant Center, where product-level markup for those fields largely repeats data Google already receives.
- Looser for custom and made-to-order catalogues, where identifiers and reviews often do not exist.
My working policy would fit on one page: zero Invalid items and zero page or feed disagreements on revenue templates; fix bad values in any field; fix missing fields the business already holds, once per template; accept and document the rest with a reason and an owner; and alert on changes in rates rather than chasing the total. A catalogue with a million warnings and that policy in place is in better shape than one with zero warnings and invented ratings.
Need help with this on your own site?
Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.
- It recommends a business-wide return policy and shipping policy under
Add a reply Cancel reply
Replies are open to approved members. Log in or apply to join.