Short answerProduct data governance assigns clear ownership, rules and routines to keep product information fit for use. It measures quality across dimensions such as completeness, accuracy, consistency, validity and timeliness, scores completeness per channel and locale, assigns accountability per attribute domain, and reviews a small KPI set in a fixed weekly, monthly and quarterly cadence.
Key takeaways
- Ownership works best at attribute-domain level, with one accountable function each for logistics, technical, marketing and compliance attributes.
- Completeness scores must be calculated per channel and per locale, because a product can be 100% complete for the webshop and 60% complete for a retailer.
- A governance cadence of weekly steward triage, a monthly data council and a quarterly data model review keeps quality from decaying after go-live.
Six quality dimensions
Product data quality becomes manageable once it is measured on defined dimensions. Each dimension needs its own rule type and its own metric.
| Dimension | Question it answers | Example rule |
|---|---|---|
| Completeness | Are all required attributes filled | Net weight required for all food items in the EU market |
| Accuracy | Does the value match reality | Dimensions match the measured physical product |
| Consistency | Do systems and channels agree | Brand name identical in ERP, PIM and webshop |
| Validity | Does the value follow format and code lists | Colour from the controlled list, units in SI |
| Timeliness | Is data available when needed | Content approved ten working days before launch |
| Uniqueness | Is each product recorded once | No duplicate GTINs across records |
Completeness and validity are easy to automate inside a PIM. Accuracy usually needs sampling or supplier certification, so organisations should budget human checks for it. Timeliness and uniqueness depend on process discipline more than tooling: launch calendars, GTIN allocation rules and duplicate checks at product creation. Measuring all six dimensions from the start sounds thorough, but it overloads teams. A sensible sequence starts with completeness and validity on priority channels, adds consistency checks between ERP and PIM in the second phase, and introduces accuracy sampling once the basics hold.
Completeness scoring
Completeness only means something per channel and per locale. A product can score 100% for the webshop and 60% for a retailer that requires extra logistics and regulatory attributes. Effective scoring follows four rules:
- Define required attribute sets per channel, locale and product family.
- Weight attributes by impact, so a missing allergen statement counts more than a missing secondary image.
- Block publication below a threshold for mandatory and regulated attributes.
- Report scores by owner, so each team sees its own backlog.
Recalculate scores automatically whenever a channel adds a required attribute, so the backlog reflects current requirements instead of last year's.
Ownership at domain level
Ownership works best per attribute domain, with a RACI that assigns one accountable function to each group of attributes. Product-level ownership collapses in catalogues above a few thousand SKUs, and system-level ownership leaves business meaning without an owner.
| Attribute domain | Accountable | Responsible | Consulted | Informed |
|---|---|---|---|---|
| Identification and logistics | Supply chain | Master data team | Purchasing | E-commerce |
| Technical specifications | Product management | Category specialists | Engineering | Marketing |
| Marketing content | Marketing | Content team | Product management | Sales |
| Compliance and safety | Regulatory affairs | Compliance specialists | Legal | All channels |
| Assets | Brand or creative | DAM librarian | Marketing | E-commerce |
Compliance attributes deserve particular attention, since requirements under the GPSR and packaging rules expand the regulated attribute set every year.
Stewardship
Data stewards turn ownership into daily practice. A steward monitors rule violations, resolves conflicts between domains, maintains code lists and trains contributors. As an estimate, organisations need one steward per 10,000 to 25,000 active SKUs, depending on attribute depth and change frequency. Stewards also own the error loop from channels: every retailer or marketplace rejection returned through syndication should land with a steward who routes it to the accountable domain.
KPIs that drive action
A short KPI set beats a long dashboard. Six indicators cover most needs:
- Completeness score per channel and locale, against target.
- Share of products blocked from publication.
- Channel rejection rate per 1,000 published items.
- Time from product creation to channel-ready status.
- Open quality issues by domain and age.
- Supplier data acceptance rate at first submission.
Each KPI needs an owner and a target that the data council reviews. Trend matters more than absolute level in the first months, because baselines are often poor. Publish the KPIs where contributors see them, for example as a dashboard in the PIM start page, so quality becomes part of daily work instead of a monthly report.
Governance cadence
Governance decays without a fixed rhythm. A workable cadence has three layers: weekly steward triage of open issues and rejections, a monthly data council where domain owners review KPIs and approve rule changes, and a quarterly review of the data model, code lists and channel requirements. Large structural changes, such as new channels or regulations, enter through the quarterly review. The same cadence prevents many implementation pitfalls that surface in the first year after go-live.
Frequently asked questions
Who should own product data quality?
Accountability should sit with business functions per attribute domain: supply chain for logistics data, product management for specifications, marketing for content and regulatory affairs for compliance. A central master data or PIM team provides stewardship, tooling and rules, while business owners decide what correct means for their attributes.
What is a good completeness target?
Mandatory and regulated attributes need a 100% target with publication blocking. For optional enrichment attributes, organisations often start between 80% and 90% for priority products and raise targets as processes mature. Targets should differ by channel, since retailer and marketplace requirements vary widely.
How many data stewards does a product catalogue need?
As an estimate, one steward per 10,000 to 25,000 active SKUs is common. Catalogues with deep technical attributes, many languages, frequent changes or heavy regulatory content sit at the lower end of that range. Automated validation rules in PIM reduce manual stewardship effort over time.