Skip to content
EncryptTechBD
Home Industries Our Work Blog Pricing About Contact Discuss Your Project
বাংলা (BN)
Data & Analytics 6 min read

Master data management for retail, explained

How retail master data management turns product, supplier and customer records into governed golden records, with a reference flow and a worked example.

E
EncryptTechBD · 0 comments
Laptop screen with analytics and a world map
Photo: Rawpixel

In a retail business, the same shirt can exist as six records. The buying team knows it by a style code, the warehouse by a carton barcode, the website by a product URL, the marketplace by a listing identifier, the point of sale by a local SKU and finance by a supplier invoice line. Each record is correct inside its own system. Together they cannot answer a simple question: how many did we sell, what did we pay and what is left?

Master data management, or MDM, is the discipline that fixes this. It is not a single purchase or a one-time clean-up. It is a governed way of creating, matching, approving and distributing the records that every other system depends on.

Why retail records disagree

Retail data fragments because channels are added faster than definitions are agreed. A new marketplace needs its own category tree. A new store opens with a local item code. A supplier changes a pack size without changing the product name. A returns team creates a new customer record because the original phone number was entered differently.

The result is not a technology failure. It is a definition and ownership failure. MDM addresses it by assigning one accountable owner per domain, one agreed definition per attribute and one process for changing either.

The domains worth governing first

  • Product: style, variant, SKU, barcode, pack size, unit of measure, category, attributes, images and status.
  • Supplier: legal entity, contacts, payment terms, lead time, currency and approved categories.
  • Customer: identity, contact details, consent, segment and linkage to orders, without duplicating records across channels.
  • Location: store, warehouse, dark store, pickup point and the identifiers each channel uses.
  • Reference data: size scales, colours, units, tax codes and category hierarchies.

Start with product and supplier data if you sell physical goods. Customer data usually follows, because it carries privacy obligations and needs a stricter consent model.

How a golden record is built

  1. Collect: ingest records from source systems, marketplaces and supplier feeds, keeping the source and timestamp.
  2. Standardise: normalise units, spellings, barcodes and category names against an agreed dictionary.
  3. Match: identify records that refer to the same entity using deterministic keys first, then fuzzy matching for names and addresses.
  4. Survive: decide which source wins for each attribute. The rule may differ by field: finance owns cost, merchandising owns the description, the warehouse owns the barcode.
  5. Approve: route exceptions to a steward who can accept, correct or reject the merge.
  6. Publish: distribute the golden record to channels through an API, event or scheduled extract, with a version history.

The survivorship rules are the heart of the design. If nobody agrees who owns an attribute, the matching engine will merely automate the argument.

Reference architecture

A practical retail MDM flow has six parts. Source systems such as the ERP, point of sale, ecommerce platform, warehouse management system and marketplace connectors feed a staging area. The staging area holds raw records and rejects. A matching service applies rules and produces candidate golden records. A stewardship workflow handles exceptions and approvals. The golden record store holds the approved version with lineage. A distribution layer publishes to channels and a monitoring layer reports match rates, completeness and change volume.

You do not need a large platform to start. A governed product table, a matching process and a publish API can deliver value before a full MDM suite is justified. What matters is that the flow is documented and that exceptions have an owner.

A worked example: one shirt, three channels

A retailer sells a cotton shirt in a store, on its website and on a marketplace. The supplier ships in cartons of six, the website sells single units and the marketplace requires a variant identifier for each size and colour.

In the source systems, the style code is the same but the SKU differs. The store uses a local code, the website uses a generated SKU and the marketplace uses its own listing ID. The barcode on the carton is not the same as the barcode on the item.

MDM creates one product record with a style, variant set, item barcode, carton barcode, pack size and category path. Each channel keeps its own identifier as a cross-reference. When stock is received, the warehouse scans the carton and the system converts it into sellable units. When an order arrives, the channel identifier resolves to the golden variant, so stock, price and returns reconcile.

The same discipline applies to the supplier record: the legal entity, payment terms and lead time are linked once, then referenced by purchase orders, landed cost calculations and performance reports.

Governance that keeps it alive

MDM decays without stewardship. Assign a data owner for each domain, a steward who handles exceptions and a change board that approves new attributes or category structures. Define service levels for new product onboarding and for correcting a wrong record. Keep a log of every change, including who approved it and which systems received it.

Make the cost of bad data visible. Count the orders held because a barcode was wrong, the returns caused by a mismatched variant and the hours spent reconciling marketplace listings. Those measures justify the stewardship time far better than a general argument about data quality.

Common misconceptions

  • MDM is a software purchase: the tool supports the process; the decisions about definitions and ownership come first.
  • It is a one-time clean-up: new products, suppliers and channels create new exceptions every week.
  • It only covers products: supplier, customer and location data usually carry equal operational risk.
  • The ERP already does it: an ERP may hold a product master, but it rarely governs marketplace, ecommerce and store identifiers in one model.
  • Perfect data is the goal: the goal is fit-for-purpose data with known exceptions and an owner for each.

When it matters

MDM becomes urgent when a retailer adds a channel, opens a new location, changes an ERP or starts reporting across banners. It is also the underlying fix for inventory synchronisation errors, duplicate customer records and supplier cost disputes. If the business runs one store, one channel and a small catalogue, a lightweight product dictionary may be enough.

Data quality measures

  • Match rate and duplicate rate by domain.
  • Attribute completeness for channel-critical fields.
  • Share of products with a verified barcode and pack size.
  • Time to onboard a new SKU or supplier.
  • Orders or listings blocked by a data exception.
  • Reconciliation breaks between physical stock and system stock.

Getting product data under control

If you need help defining product and supplier records before a channel or ERP change, start with Business Automation. The pricing page explains how data assessment, implementation and ongoing stewardship are scoped. For the channel operations that depend on this data, read how omnichannel retail operations change year by year.

Sources

Services that deliver this

Want this implemented properly?

Pricing →

Every engagement starts with a written scope. If a requirement is outside it, we quote it separately before starting.

Responses (0)

No comments yet. Add the first one below.

Leave a response

Your email address will not be published. Required fields are marked *.

Keep reading

More field notes