Product data quality is easiest to improve when the team can see it as a measurable operating system. A scorecard turns vague complaints such as “the catalog is messy” into a shared view of which products, attributes, suppliers, and channels need attention.
The six dimensions of product data quality
A useful scorecard separates different failure modes. A product can be complete but inaccurate, accurate but inconsistent, or rich in content but not ready for a specific channel.
- Completeness: Are the required fields populated for the category and channel?
- Accuracy: Do values match the source of truth, product specification, and approved business rules?
- Consistency: Do units, names, formats, and values follow the same conventions?
- Uniqueness: Are duplicate products, SKUs, variants, and assets under control?
- Freshness: Are price, availability, compliance, and other time-sensitive fields current?
- Channel readiness: Can the product pass the requirements of each destination without manual repair?
A practical weighting model
Start with a 100-point score. Weight completeness at 25 points, accuracy at 20, consistency at 15, uniqueness at 10, freshness at 10, and channel readiness at 20. A regulated or marketplace-heavy catalog may increase the weight of accuracy and channel readiness.
Score at the product level, then aggregate by category, supplier, brand, channel, and market. The aggregate score is useful for leadership; the underlying exceptions are what operators need to fix.
Suggested scorecard fields
| Dimension | Example checks | Evidence |
|---|---|---|
| Completeness | Title, description, category, price, availability, images, required attributes | Required-field rules by category |
| Accuracy | Specification, dimensions, identifiers, claims, price, stock | Source record or approved reference |
| Consistency | Units, casing, colour names, naming patterns, decimal formats | Controlled vocabulary and normalization rules |
| Uniqueness | SKU, GTIN, variant and image duplication | Duplicate report and merge decision |
| Freshness | Last update, stock age, price age, expiring compliance values | Timestamp and source-system status |
| Channel readiness | Category mapping, identifiers, image rules, policy fields, landing page | Channel validation result |
Set thresholds that trigger action
Use simple thresholds at first. For example, a product can be green at 90–100, amber at 75–89, and red below 75. Add hard stops for critical fields: a missing price, availability, legal claim, or channel-required identifier may block publication even when the overall score is high.
Do not use one threshold for every category. A fashion product and an industrial component have different required attributes. Rules should inherit from category and channel requirements.
Move from score to workflow
Every red result should become an actionable exception with an owner. Group repeated errors by root cause: one supplier sending the wrong unit, one mapping that creates empty categories, or one import that strips variant values. Fixing the root cause improves many products at once.
Compare the scorecard with the product data quality checklist and use the PIM data quality guide to define a repeatable review cadence.
A worked example: scoring one product
Take a single product, a dining chair, and apply the 100-point model from above.
| Dimension | Max | Result | Score | Reason |
|---|---|---|---|---|
| Completeness | 25 | 9 of 10 required fields | 22.5 | Assembly instructions missing |
| Accuracy | 20 | Width does not match spec sheet | 12 | Unit conversion error |
| Consistency | 15 | Colour uses a non-standard value | 10 | “Dk Grn” instead of “Dark green” |
| Uniqueness | 10 | No duplicates | 10 | Passed |
| Freshness | 10 | Stock updated yesterday | 10 | Passed |
| Channel readiness | 20 | Category mapped, GTIN missing | 10 | Blocks marketplace publishing |
| Total | 100 | 74.5 | Red (below 75) |
This product scores 74.5, so it is red. The missing GTIN is a hard stop regardless of the total. The fixes are specific: correct the width, normalize the colour, add the GTIN and the assembly document.
How to sample when the catalog is large
You do not need to score every product on day one. Score a representative sample, then expand.
- Segment by category, supplier and channel.
- Take the top-selling and highest-traffic products first, since errors there cost the most.
- Add a random sample from the long tail to detect systemic problems.
- Re-run the same sample after fixes so before and after are comparable.
Roll up the score for different audiences
| Audience | View | Use |
|---|---|---|
| Leadership | Overall score and trend by category | Prioritize investment |
| Category managers | Score by category with top exceptions | Plan enrichment work |
| Suppliers | Score for their products and common errors | Improve incoming data |
| Channel owners | Channel readiness and rejection reasons | Reduce feed errors |
| Data stewards | Item-level exceptions queue | Fix and approve |
Cadence and ownership
A scorecard only works if someone owns it. Assign a data steward per category, review the score weekly for operational exceptions and monthly for trends, and run a deeper review each quarter or after a major import. Record the thresholds and weights in the governance documentation so changes are deliberate. See product data governance roles and approval workflows for how to set this up.
Common scorecard mistakes
- Averaging away critical errors. A high average can hide a missing price. Use hard stops.
- One rule set for every category. Required fields differ by category and channel.
- Measuring without fixing. Every red item needs an owner and a due date.
- Changing the formula too often. Keep weights stable so trends mean something.
- Ignoring the source. If one supplier causes most errors, fix intake rather than cleaning every file.
Scorecard FAQ
What is a good product data quality score?
There is no universal number. Set your own thresholds, for example green at 90 or above, and focus on moving high-impact products first.
How often should I update the scorecard?
Automate it where you can, so it refreshes whenever data changes. Review exceptions weekly and trends monthly.
Can I build this in a spreadsheet?
Yes, for a first audit. Move to automated rules once you need repeatable, per-channel validation.
Choosing the rules behind each dimension
A scorecard is only as good as the rules it runs. Write each rule as a testable statement, so it can be automated later.
| Dimension | Example rule | Severity |
|---|---|---|
| Completeness | Title, description, category, price and at least one image are present | Hard stop |
| Completeness | All category-required attributes are filled | Weighted |
| Accuracy | Dimensions fall within a plausible range for the category | Weighted |
| Accuracy | GTIN passes the check-digit test | Hard stop for channels that require it |
| Consistency | Colour values come from the controlled list | Weighted |
| Uniqueness | No two active products share a GTIN or SKU | Hard stop |
| Freshness | Price was updated within the agreed window | Weighted |
| Channel readiness | Mapped category and all channel-required fields exist | Hard stop per channel |
Run identifier rules against the GTIN Validator when you first build the rule list, and read the guide to GTIN check digits for the logic behind it.
Where the data comes from
List the source of truth for each field: the ERP for price and stock, the supplier file for specifications, the PIM for enriched content, and the commerce platform for what is live. A scorecard that compares systems can uncover mismatches that no single system shows. Record the source in the rule definition so a failing check points to the system that needs correcting.
Start small and expand
Start with five to ten rules that cover your biggest known problems, such as missing images, missing identifiers and inconsistent colour values. Add rules as you learn which errors cost the most, and retire rules that never fail.
Measure business impact
Connect quality scores to outcomes such as search exits, zero-result searches, feed disapprovals, returns, support contacts, conversion rate, and revenue per session. The goal is not a perfect number for its own sake. The goal is to show which quality improvements make the catalog easier to discover, trust, compare, and buy.
Start with a sample of important products, establish a baseline, fix the highest-impact issues, and measure again after publication. The Completeness Checker can turn the first manual audit into a repeatable validation step.
Product data quality scorecard with validated catalog records and quality checks
