Skip to content
Tutorials Center
My learning

Topics

Lesson 24 of 36 ยท Integration and Data

Master Data Management in ERP and CRM

Understand master data vs transactional data, golden records, ownership and the duplicate problem, and try matching and merging customer records with a Python stand-in.

Every business process touches the same few kinds of information over and over: who the customers are, which products exist, what suppliers charge, how the company is structured. This slow-changing, shared information is master data. When master data is wrong, every process built on it is wrong too: invoices go to the wrong address, stock is counted twice under two part numbers, and marketing emails the same person three times.

Master data management (MDM) is the set of rules, roles and tools that keep master data consistent, complete and trusted across ERP, CRM and other systems. The product is not available on this site, so the examples use Python to show the core techniques.

Master data vs transactional data

Master data Transactional data
Describes Business objects (customer, product, supplier, employee, account) Business events (order, invoice, payment, call)
Change rate Slow Constant
Volume Moderate Very large
Lifetime Years Documents are archived
Example Customer 4711, address, tax ID Sales order 90001 for customer 4711

There is also reference data (country codes, currencies, units of measure, payment terms), small lists shared by many records, and metadata (definitions of fields and rules).

Typical master data domains: customer, supplier, material/product, chart of accounts, employee, asset, bill of materials and price lists.

One record, many systems

In a growing company the same customer lives in the CRM (as an account with contacts), the ERP (as a sold-to, ship-to and bill-to), the e-commerce shop and the support tool. Each system holds a slightly different version. MDM answers three questions: who owns each attribute, which system is the source of truth, and how are changes distributed.

MDM styles

Registry Index of IDs; data stays in each system.
Coexistence Changes flow both ways.
Central hub One system creates and owns the record.

Most companies mix styles by domain.

1 / 2
  • Registry: a central index links records across systems with a global ID. Little data moves, but each system must be consulted to read the full picture.
  • Consolidation: data is copied to a hub for analysis, cleaned there, but not written back.
  • Coexistence: the hub holds the golden record and synchronizes changes both ways.
  • Centralized (transaction) hub: records are created and changed only in the hub, and other systems receive them. Strongest control, most change to processes.

The golden record

A golden record is the best, most complete and most current version of a business entity, built from several source records. Building one uses three steps:

  1. Standardize: normalize case, whitespace, punctuation, country and phone formats, legal-form suffixes ("Ltd", "Limited").
  2. Match: find records that describe the same real-world entity (exact keys, such as tax ID, or fuzzy comparison of name and address).
  3. Merge (survivorship): for each attribute choose which value wins: the most recent, the one from the most trusted system, or the most complete.
Python Standardize and match customer records

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Matching always has a gray zone. Pairs above a high threshold merge automatically; pairs in between go to a data steward for manual review; pairs below are treated as different. Raising the threshold lowers wrong merges (false positives) but leaves more duplicates (false negatives). Wrong merges are usually worse, because they mix two customers' invoices.

Survivorship rules

Once a group of records is known to be one customer, build the golden record attribute by attribute.

Python Merge duplicates into a golden record

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Keep lineage: store which source each value came from, so a steward can explain the result and undo a bad merge.

Governance: roles and processes

Technology alone does not fix master data. A working MDM programme defines:

  • Data owners (business executives accountable for a domain, such as the head of sales for customers).
  • Data stewards (hands-on people who resolve duplicates, approve new records and fix errors).
  • Data standards (mandatory fields, formats, naming rules, allowed values).
  • Creation and change workflows (who may request, who approves, how fast).
  • Quality metrics and reviews (see data quality monitoring).

A new customer should be created once, by a controlled process, with checks for duplicates before saving. Prevention is much cheaper than cleanup (see data migration and cleansing for the one-time cleanup).

Common mistakes

  • Treating MDM as an IT project with no business owner.
  • Merging automatically at a low threshold and creating wrong records.
  • Letting each system keep its own customer number with no cross-reference.
  • Allowing free-text values for things that should be reference data (country, industry).
  • Deleting superseded records instead of marking them inactive and keeping history.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. Which of these is master data?

    Master data describes shared business objects such as customers, products and suppliers. Orders, postings and calls are transactions.
  2. 2. What is a golden record?

    A golden record combines the best attributes from all duplicates using survivorship rules.
  3. 3. Why are wrong merges usually worse than missed duplicates?

    A duplicate is annoying but harmless; a wrong merge corrupts business transactions, so automatic thresholds are set conservatively.
  4. 4. Who is accountable for the quality of customer master data?

    Ownership must sit with the business. Stewards handle the daily work.

Summary

  • Master data describes shared business objects; transactional data records events.
  • MDM styles range from a registry to a central hub; most firms mix them.
  • A golden record comes from standardizing, matching and merging with survivorship rules.
  • Use thresholds with a manual review zone; keep lineage.
  • Owners, stewards and creation workflows matter more than the tool.

Next: Data Migration and Cleansing