# Governing product, location, and customer master data
Two grocery banners merge. Eighteen months later, finance still cannot answer a simple question: how many stores does the combined company actually operate. Store 0412 exists twice, once under the acquirer's numbering and once under the legacy system, both still transacting. Nobody decommissioned the duplicate. Every regional sales report since the merger has been double-counting that location's revenue by a few percentage points, and nobody noticed because the number "looked plausible."
This is not a hypothetical edge case. It is the default outcome when master 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 → is treated as an IT afterthought instead of a business discipline. This lesson covers the core master data domains in retail, the quality metrics that catch problems like the one above, and the governance mechanisms that prevent them.
Master data is the core, relatively stable reference data that every transaction depends on. It is distinct from transactional data (a sale, a shipment, a return), which is generated constantly and references master data records.
The three domains that matter most in retail and distribution:
Product master data: SKU (stock keeping unit, a unique code for a sellable item) attributes such as description, category hierarchy, size, UPC/EAN barcode (Universal Product Code / European Article Number), unit of measure, and supplier link.
Location master data: store IDs, distribution center codes, banner/format flags, geographic attributes, square footage, open/close dates.
Customer master data: loyalty program IDs, household linkage, contact and consent records, purchase history keys.
A fourth, often underestimated, domain is vendor/supplier master data: the entities you buy from, which link to cost, lead time, and compliance data.
Each domain typically lives in a "system of record", the authoritative source, but gets copied into a dozen downstream systems: point-of-sale (POS), the data warehousedata warehouseA central repository that consolidates data from many source systems into a structured, query-optimized store designed for analytics, reporting, and business intelligence.Voir la définition complète →, the e-commerce platform, the planning tool, the loyalty engine. Every copy is a chance for drift.
A single-banner retailer can survive sloppy master data for years because everyone unconsciously learns the workarounds. A merger removes that tribal knowledge overnight.
Common failure modes:
These are silent failures. Reports still run. Dashboards still populate. Numbers just quietly drift from truth, often for months, because nothing crashes.
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 → is usually assessed across a standard set of dimensions, borrowed from the broader data management field and codified in frameworks like DAMA International's Data Management Body of Knowledge (DMBOK):
Retail data teams typically monitor a small set of concrete, auditable metrics rather than abstract quality scores.
Duplicate rate: percentage of records in a master table that are duplicates of another record.
*Worked example*: a location master table has 3,200 rows after a merger. A matching exercise (comparing address, tax ID, and geocode) identifies 140 duplicate pairs.
Duplicate rate = 140 / 3,200 = 4.4%
That means roughly 1 in 23 "stores" in the system is a phantom, inflating footprint counts and double-counting revenue in any report that sums by location.
Attribute completeness rate: percentage of mandatory fields populated.
Example: if a product table requires category, sub-category, UPC, and unit of measure, and 92,000 of 100,000 SKUs have all four filled, completeness = 92%.
Cross-system consistency rate: percentage of records where a given attribute matches across two systems (e.g., POS category vs. warehouse management system category). Retailers running post-merger harmonization projects often start below 70% and target above 98% before decommissioning legacy systems.
Golden record coverage: percentage of customer transactions successfully linked to a single deduplicated customer profile, rather than orphaned or split across multiple IDs.
Time-to-correct: average time between a 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 → issue being flagged and being resolved. A mature governance function tracks this like an incident SLA (service level agreement).
These are estimates of typical target ranges cited in retail data management practice as of the mid-2020s, not universal standards; actual thresholds should be set per company based on business risk.
Metrics tell you there is a fire. Governance mechanisms stop it from starting.
1. Single system of record per domain. Decide, explicitly, which system owns the truth for location, product, and customer. Everything else subscribes to it, it does not maintain its own version.
2. Master Data Management (MDM) tooling. Platforms (examples include Informatica MDMMDMMaster Data Management (MDM) is the discipline of creating and maintaining a single, consistent, trusted version of an organization's core business entities like customers, products, and suppliers.Voir la définition complète →, SAP Master 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 →, and Reltio) that centralize matching, deduplication, and distribution of golden records to downstream systems. In a merger, MDMMDMMaster Data Management (MDM) is the discipline of creating and maintaining a single, consistent, trusted version of an organization's core business entities like customers, products, and suppliers. tooling is where the crosswalk tables between legacy and new IDs live.
3. Data stewardship roles. A named business owner (not just IT) accountable for each domain's quality: a merchandising lead owns product hierarchy rules, a real estate/operations lead owns location data.
4. Change control on hierarchies. Category and location structures should not change without a governed approval process, because every historical report implicitly depends on the hierarchy staying stable (or being versioned).
5. Automated validation rules at entry. Reject a new SKU record missing a UPC. Flag a new store record with an address that fails geocoding. Catching errors at creation is far cheaper than cleaning them up 18 months later.
A simple validation rule, expressed in pseudocode, illustrates the entry-point check:
IF new_location.store_id EXISTS in location_master
AND new_location.address MATCHES existing.address
THEN flag_as_potential_duplicate()
ELSE IF new_location.upc_length != 12
THEN reject("invalid UPC format")
ELSE approve_and_publish()Vérification des acquis
1. What best distinguishes master data from transactional data in a retail context?
2. In the merger scenario, Store 0412 was duplicated under two different numbering systems and both records kept transacting. What is the root cause this illustrates?
3. A company wants a single authoritative source for each product's attributes, even though the data is copied into a dozen downstream systems. What concept describes that authoritative source?
4. Select ALL correct answers: which of the following are examples of location master data attributes, as distinct from product or customer master data?
Sélectionnez toutes les réponses correctes.
5. Select ALL correct answers: why did the duplicate-store problem in the merger go unnoticed for eighteen months?
Sélectionnez toutes les réponses correctes.
There is no single global "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" published for retail the way there is, say, an inflation index. What exists are internal maturity assessments and vendor-published survey estimates (treat these as directional, not precise): industry surveys on 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 → (from firms like Gartner and various MDMMDM vendors) have for years estimated that organizations believe poor costs them significant sums annually, though exact dollar figures vary widely by methodology and should not be quoted as fixed facts.
The more actionable approach for a retail data team: build an internal quality scorecard per domain (product, location, customer, vendor), track the metrics above monthly, and set your own improvement targets tied to specific business consequences, like the merger reporting example, rather than chasing an external benchmark number.
For a broader grounding in 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 → management principles, DAMA International's DMBOK overview is a solid, freely accessible reference point.