Content health score
A content health score is a composite grade, usually 0-100, that rates how complete and compliant a product listing is against a rubric. Typical checks cover required attributes, image specs, description length, and identifiers. Retailers, PIM vendors, and analytics tools each write their own rubric, so the same SKU can pass in one system and fail in another. The score measures conformance to a checklist, not whether a buyer can choose the product.
What the score actually measures
Every content health score is two things: a rubric of checks, and a weighting that decides how much each check moves the number. Strip away the dashboard and you are looking at a spreadsheet of pass/fail tests rolled up into one figure.
Most rubrics draw from four families of checks.
| Check family | What it looks at | What a failure looks like |
|---|---|---|
| Completeness | Required and recommended attributes populated for the category | A 3/8-16 x 2 in. Grade 8 hex bolt with no thread pitch, no finish, and no proof load rating |
| Identifiers | GTIN, UPC, MPN, brand, and their check-digit validity | A GTIN-14 that fails its check digit, or an MPN copied from the distributor's internal part number |
| Media | Image count, minimum pixel dimensions, white-background primary, video, spec sheet PDF | One 480x480 supplier thumbnail where the retailer requires six images at 1600px |
| Copy | Title format, bullet count, description word count, banned-claim scan | A title reading "WIRE CONN 600V UL" where the spec wants brand, type, gauge range, and voltage |
Weighting is where scores diverge. A rubric that gives images 40% of the total will punish a well-specified fastener with one photo harder than a rubric that gives images 10%. The underlying product is identical in both cases; only the weighting differs.
Why the same SKU scores differently in two tools
Three groups publish content health scores, and they are answering different questions.
- Retailers and marketplaces score against their own listing spec. The question is "does this meet our requirements for this category." Two retailers selling the same UL listed 600V wire connector can publish different required-attribute sets for it, so identical data clears one and fails the other.
- Digital shelf analytics tools score what they can scrape from the live PDP. The question is "how does this listing compare to competitors on the same search result." They see rendered output, not your PIM record.
- PIM and syndication platforms score the record you own. The question is "is this channel-ready before we publish."
The practical consequence: a score is only meaningful next to its rubric. "We're at 78" describes one grader's rubric, not the state of your catalog.
Before you act on a score, get three things in writing from whoever produced it: the check list, the weights, and the category scope. Without those, you cannot tell an easy 12-point gain from an impossible one.
What the score does not tell you
Content health scores measure checklist conformance. They are blind to several things that decide whether a SKU actually sells.
Correctness is the largest gap. A rubric checks that the grade field on a hex bolt is populated; it does not check that the value is true. A 3/8-16 x 2 in. bolt that is actually Grade 5 can be published as Grade 8 with every required field filled, six images at the right resolution, a valid GTIN, and a title in the retailer's exact format. That listing reads 100. It is also wrong in the one way that gets a fastener returned or an assembly re-inspected. Validating the value against a mill certificate or a manufacturer spec sheet is work no rubric performs.
The remaining blind spots:
- Findability. Filling an attribute does not put it on the facet. If thread pitch is stored as free text rather than a controlled value, the listing takes full marks and still misses the filter.
- Buyer questions. No rubric asks whether the page answers "will this fit my existing panel."
- Cost to fix. A 3-point gap from a missing video and a 3-point gap from missing UL file numbers across 40,000 SKUs are the same number and wildly different projects.
That last one is the trap in reporting the score upward. It moves when you add bullet points. It also moves when someone has to find the UL file number on a supplier's spec sheet, which is the same three points at a completely different cost.
Score versus fill rate
A score becomes useful when you decompose it back into the failing checks and sort those by cost and impact.
| Instead of | Do this |
|---|---|
| "Catalog health is 74" | Name the failing checks: these SKUs are missing a compliance rating, these SKUs are missing a primary image |
| Chasing the average up | Fixing the checks that gate a specific channel's ingestion |
| One score per catalog | One score per category, because rubrics are written per category |
| Reporting the score monthly | Reporting fill rate on the attributes buyers actually filter on |
Attribute fill rate is the more honest primitive. It is a single unweighted number you can argue with, and it maps directly to work: this attribute, these SKUs.
A score reports which fields are empty; sourcing a verified value for each one is a separate exercise.
Frequently asked questions
What is a good content health score?
There is no portable answer, because the number depends entirely on the rubric and its weights. A 90 against a five-check internal rubric is easier than a 70 against a retailer's full category spec. Rather than targeting a number, target the checks that gate something concrete: channel ingestion, a facet you want to appear on, or a compliance field your legal team requires.
How is a content health score different from attribute fill rate?
Fill rate is a single measurement: what percentage of SKUs have a value in a given attribute. A content health score is a weighted composite of many checks, including fill rate, image specs, copy length, and identifier validity. Fill rate is harder to game and easier to act on. Health scores are better for executive reporting and worse for planning the work.
Do retailers actually use content health scores to rank products?
Treat a specific retailer's score as a compliance gate you have to clear rather than a ranking lever you can push. Ranking behavior is not published in enough detail to reverse-engineer from a score anyway. What is verifiable without any score: incomplete records fail ingestion, drop off facets, and cannot be filtered into. That is reason enough to fix the failing checks.
Can a listing score 100 and still be wrong?
Yes, and this is the main limitation. Rubrics check presence and format, not truth. A hex bolt listed with the wrong grade, a connector with an unverified UL claim, or a mismatched MPN all pass a completeness check cleanly. Correctness requires validating each value against a source document, which is a different exercise from scoring.
Doesn't our PIM already give us a content health score?
If your PIM reports a completeness or readiness score, that tells you which fields are empty against the attributes you configured. It does not fill them. The distance between knowing a field is missing and having a sourced, verified value in it is where the real cost of catalog work sits.