Data quality metrics for product and material master data

# Data qualityData qualityThe degree to which data is fit for purpose: accurate, complete, consistent, timely, valid and unique. Poor quality data undermines analytics, reporting and AI.Voir la définition complète → metrics for product and material master data

A leather goods merchandiser at a European maison pulls the SKU (Stock Keeping Unit, a unique code identifying each distinct product variant) list for a bestselling tote and finds three different color codes for the same "Nero" black, two conflicting leather-type fields, and a material cost that has not updated in fourteen months. The e-commerce team, working off a different extract, lists the bag as available in a colorway that manufacturing discontinued last season. Nobody lied. The data just quietly rotted.

This is the ordinary condition of product master data in luxury. Small batch sizes, frequent limited editions, multiple ERP (Enterprise Resource Planning, the core system managing inventory, production and orders) migrations after acquisitions, and hand-entered attributes at ateliers all combine to produce catalogs that look clean in a slide deck and fall apart at the SKU level.

Why product and material master data is uniquely fragile in luxury

Mass-market retailers deal with tens of thousands of SKUs from a handful of factories. A luxury house can run thousands of SKUs per season, many produced in runs under 500 units, sourced from dozens of tanneries and ateliers, each with its own naming conventions.

Master data here means the core reference records that other systems depend on: product ID, material composition, colorway, size run, supplier code, HS code (Harmonized System code, the international customs classification used for cross-border shipping), and price. When these records diverge across PLM (Product Lifecycle Management, the system tracking a product from design to discontinuation), ERP, and e-commerce platforms, everything downstream breaks: sizing charts, customs declarations, sustainability disclosures, and personalization algorithms.

Material master data is the sharper pain point. A single handbag might reference calfskin, brass hardware, and cotton lining, each needing its own traceability record for regulations like the EU's forthcoming Digital Product Passport (a mandatory digital record of a product's materials and origin, phased in from 2027 under the EU Ecodesign for Sustainable Products Regulation, ESPR). Get the material code wrong and the compliance document is wrong too.

The core datasets to know

• PLM records: design specs, bill of materials, seasonal colorways, sample approvals.
• ERP/inventory data: stock levels, warehouse locations, SKU status (active, discontinued, archived).
• Supplier and material master files: tannery codes, fabric composition, certifications (e.g., Leather Working Group ratings).
• E-commerce product catalog: customer-facing descriptions, images, size guides, pricing by market.
• Customs and trade data: HS codes, country of origin, used for cross-border compliance.

Each of these is typically owned by a different team, updated on a different cadence, and reconciled (if at all) manually. This is the root cause of most data-quality failures in the sector.

The three metrics that matter most

Completeness

Completeness measures the share of required fields that are actually populated. For a leather goods SKU, a minimal required field set might be: product ID, material composition, color code, size, weight, HS code, and supplier ID.

Worked example: Say a catalog has 4,000 active SKUs, and each requires 8 mandatory fields (32,000 total field-instances). An audit finds 2,240 fields blank or filled with placeholder text like "TBD."

Completeness rate = (32,000, 2,240) / 32,000 = 0.93, or 93%.

An estimate commonly cited in retail data-quality benchmarking (see DAMA International's Data Management Body of Knowledge for the general framework) treats anything below 95% completeness on mandatory fields as a governance red flag requiring remediation before the data feeds customer-facing or compliance systems.

Consistency

Consistency measures whether the same real-world entity is described the same way across systems. The "Nero" black tote with three color codes is a consistency failure, not a completeness one: every field is filled, just contradictorily.

A simple consistency check: pick a sample of SKUs that exist in both PLM and e-commerce, and calculate the percentage where key attributes (material, color, size) match exactly.

Consistency rate = (SKUs with matching attributes across systems) / (total SKUs sampled)

If 500 SKUs are sampled and 460 match cleanly, consistency = 92%. Luxury houses running multiple legacy ERPs after mergers often see this figure drop into the 70 to 85% range (estimate, based on typical post-merger system integration patterns), which is precisely why integration projects budget heavily for data cleansing, not just software licensing.

Freshness (or timeliness)

Freshness measures how current a record is relative to when it should have been updated. A discontinued colorway still marked "in stock" online is a freshness failure.

A workable metric: average age of "stale" records, defined as records untouched since a triggering event (season close, price change, discontinuation) that should have updated them.

If discontinuation is flagged in PLM on day 0 but the e-commerce catalog is only synced weekly, an acceptable freshness SLA (Service Level Agreement, a defined performance target) might be "under 7 days lag." Anything beyond that window on customer-facing channels risks selling products that no longer exist, a direct driver of returns and customer complaints.

Turning metrics into a scorecard

A simple governance dashboard combines the three into a composite Data QualityData QualityThe degree to which data is fit for purpose: accurate, complete, consistent, timely, valid and unique. Poor quality data undermines analytics, reporting and AI.Voir la définition complète → Index (DQI), often weighted by business impact:

DQI = (0.4 × Completeness) + (0.35 × Consistency) + (0.25 × Freshness)

Weights are a judgment call, not a universal standard, calibrated to what breaks the business most. A house selling heavily through its own e-commerce site might weight freshness higher; a house focused on wholesale and customs compliance might weight completeness of HS codes and material composition higher.

Applying illustrative numbers (93% completeness, 92% consistency, 88% freshness):

DQI = (0.4 × 93) + (0.35 × 92) + (0.25 × 88) = 37.2 + 32.2 + 22 = 91.4

This single number becomes trackable over time, per product line, and per data source, giving merchandising and IT a shared target instead of separate anecdotes.

Governance: who owns the fix

Metrics without ownership decay again within a season. Leading practice, drawn from general data governancedata governanceData governance is the set of policies, roles, and processes that ensure data is accurate, secure, well-defined, and used responsibly across an organization.Voir la définition complète → frameworks like DAMA-DMBOKDAMA-DMBOKData Management Body of Knowledge, référentiel de l'association DAMA définissant les 11 domaines de gestion des données (gouvernance, qualité, architecture, sécurité, etc.)., assigns:

• Data stewards per domain (one for product, one for materials, one for supplier data) responsible for defining "correct" values and resolving conflicts.
• A single source of truth for each attribute type, usually PLM for design attributes and ERP for inventory and logistics, with e-commerce as a downstream consumer, never an originator.
• Change triggers: any colorway discontinuation, price change, or material substitution must fire an update event to all downstream systems, not rely on the next manual export.