Skip to content

ERP · MDM

MDM — one trusted record of everything

Master data management: a single, governed version of your customers, vendors, products and materials across every system. The unglamorous work that decides whether your reporting is trusted and whether an AI agent can be let near your operations.

Sound familiar?

  • The same customer exists four times across your CRM, ERP and billing system, with three different spellings.

  • Two teams produce different revenue numbers from the same month and nobody can say which is right.

  • You want to automate a process, but nobody trusts the underlying data enough to let software act on it.

What we do

Capabilities

Data audit
Where your master data actually lives, how much of it conflicts, and what that is costing in rework and wrong decisions.
Domain modelling
Agreeing what a customer, vendor, product or material is — the definitional work that most MDM programmes skip and then fail on.
Deduplication and cleansing
Matching, merging and surviving records, with the rules documented so the result is reproducible rather than a one-off cleanup.
Governance and stewardship
Who owns which fields, who approves changes, and what happens when two systems disagree. Without this, data degrades again within a year.
Golden record and distribution
A single authoritative record, published to the systems that need it, so the ERP and the CRM stop drifting apart.
AI readiness
The data quality floor an agent needs before it can be trusted to read from, let alone write to, your operational systems.

Outcomes

What changes

  • Before

    The same customer exists in three systems under three spellings.

    After

    One governed record, distributed, with each system still able to recognise it.

  • Before

    Master data is created by anyone with access, straight into the ERP.

    After

    Every record arrives through a request that was validated and approved by its owners.

  • Before

    Reporting disputes are settled by deciding whose extract to believe.

    After

    There is one record to believe, so the disagreement moves on to what it means.

How it works

The engagement

  1. 01

    Audit

    2 weeks

    Profiling across your systems to quantify duplication, conflict and completeness. You get the numbers before you commit to a programme.

  2. 02

    Model and govern

    3–4 weeks

    Domain definitions, field ownership, survivorship rules and the approval workflow. The definitional decisions, agreed with the people who own the data.

  3. 03

    Cleanse and consolidate

    4–8 weeks

    Matching and merging against the agreed rules, with a reviewed exception queue rather than silent automatic merges.

  4. 04

    Distribute and sustain

    ongoing

    Publishing the golden record to downstream systems and monitoring quality so it does not quietly degrade.

Go deeper

The playbooks set out how these processes run, step by step — what is standard, what gets configured, and what changes.

Read the playbooks

Stack

  • SAP Master Data Governance
  • Odoo data models
  • dbt
  • Great Expectations
  • Python
  • PostgreSQL

Questions

The things people actually ask

Is MDM not just data cleanup?
Cleanup is a one-off; MDM is the governance that stops it being necessary again. If you only cleanse, you will be back where you started within about a year, which is the most common way these projects waste money.
Do we need a dedicated MDM tool?
Often not. For companies of your size, the governance model matters far more than the software, and the ERP plus a well-designed pipeline usually suffices. Dedicated tooling earns its cost at higher record volumes or with strict regulatory lineage requirements.
Can we do this at the same time as the ERP implementation?
Yes, and it is the best time. Migration already forces you to decide what a customer is — doing the governance work then costs far less than retrofitting it afterwards.
Who needs to be involved from our side?
The people who own the data, not just IT. Typically finance, sales operations and whoever runs your supply chain. MDM fails when it is treated as a technical project rather than a set of business decisions.
How does this connect to the AI work?
Directly. An agent reading from duplicated or conflicting master data produces confident, wrong answers, and an agent writing into it makes the problem worse. MDM is the floor that makes the agent layer safe.
What does success look like?
Two teams pulling the same number from different systems and getting the same answer, without anyone reconciling by hand.

Tell us what’s breaking.

Thirty minutes, no deck. We’ll tell you whether this is the right place to start.

Book a 30-minute call