Data strategy and data governance
Master data: the unglamorous foundation of every AI use case
Before you ask what AI can do with your data, ask whether your data knows who your customers are.
Ask a leadership team about its AI ambitions and you will hear about forecasting, churn prediction and AI agents handling customer queries. Ask how many customers the company has, and you may get three different answers. That second question matters more than it seems.
Why AI makes the problem bigger
A report built on messy data is visibly wrong. A sales manager spots a duplicate customer and works around it. A model that learns from duplicate or inconsistent records can't: it learns the wrong patterns and then presents them with confidence. An AI agent that pulls a customer's history from three conflicting records gives an answer that looks right and still isn't. With AI, poor master data becomes a bigger problem.
What good master data looks like
Master data is the shared reference information the business runs on: customers, products, suppliers, the chart of accounts. It starts with one definition. What is a "customer": a legal entity, a site, a billing account? This is where most of the confusion begins. Then every customer, product and supplier exists once, as a golden record with a unique ID that every system uses or refers to.
Each data domain also has a business owner who decides on definitions and quality rules, and data stewards who keep it in order day to day. And quality is visible: a handful of measures, such as duplicates, missing key fields and records without an owner, that anyone can look at.
Getting started without a multi-year programme
Master data projects have a reputation for being slow and expensive. They don't have to be.
- Start with a use case. Pick one AI or analytics application that matters, such as a revenue forecast or a single customer view for sales, and fix the data it needs first.
- Fix it at the source. Cleaning in the data warehouse helps for a while. Lasting quality comes from correcting data where it is created, with validation in the CRM or ERP.
- Consolidate step by step. Companies with several ERP or CRM systems rarely need a big-bang migration. A shared customer ID and a mapping layer already provide a single source of truth for reporting and AI. At Solmax, that is how we brought reporting from five ERP systems together in one governed data warehouse.
- Make quality routine. A monthly data quality review with the business owners stops it from slipping back.
Master data will never be the exciting part of an AI strategy, but it does decide whether that strategy delivers. The organisations that quietly invest in it now will be the ones whose AI can be trusted.
Read next
Let's talk
Would you like to exchange ideas on data, analytics, AI or FP&A? I would be glad to connect.