Master data: fix it at the source

Ask any finance or operations team why two reports disagree, and the answer usually comes back to master data. The same customer exists three times under slightly different names. A product is coded one way in the warehouse and another in sales. A supplier has no payment terms recorded, so someone checks a folder every time an invoice arrives.

Duplicate customers, inconsistent items and missing fields cause most reporting disputes. They also slow down everyday work, from order entry to month-end close. Many organisations respond with periodic clean-up projects that fix the data and then watch it decay again. This guide explains how to fix it at the source, so the improvement lasts.

What master data is and why it matters

Master data is the shared reference information that transactions depend on: customers, suppliers, products, materials, employees, locations and the chart of accounts. It changes relatively slowly, but every order, invoice, shipment and journal uses it. An error in a customer record is repeated in every transaction with that customer.

• Duplicate customers split sales history, so account managers and credit control each see part of the picture.
• Inconsistent product codes and units make stock figures unreliable and cause picking errors.
• Missing supplier details delay payments and create manual checks.
• Wrong cost centres or account mappings distort management reports.

None of this is a technology problem at heart. Most business systems can hold good master data. The problems come from how data is created and maintained: who can add a record, what they must fill in, and whether anyone notices when it is wrong.

Name an owner

Each master data set needs a person accountable for quality. Without an owner, everyone can create and change records and nobody is responsible for the result. With one, there is someone who decides the rules, monitors quality and resolves disputes.

Ownership is often contested. Sales sees customer records as its own, finance needs them for credit and invoicing, and logistics relies on delivery addresses. That is exactly why one owner is needed. The owner does not do all the work; they set the rules everyone follows and make the final call when departments disagree.

The owner should usually sit in the business, close to the data’s use. Credit control or sales operations often own customer data. Purchasing or accounts payable own supplier data. A product or supply chain manager owns item data. IT supports them with tools and access, but does not decide what a correct record looks like.

What the owner does

1. Defines what a complete, correct record looks like, including required fields and formats.
2. Decides who may create and change records, and how requests are made.
3. Monitors quality with a few simple measures, such as duplicates found and incomplete records.
4. Resolves disputes, for example when sales and finance disagree on a customer’s structure.
5. Reports on quality regularly to the relevant leadership.

Prevent bad entries

Use validation rules and required fields at entry. The cheapest error to fix is the one that is never created. Most business systems allow rules that stop or warn about incomplete or inconsistent records at the moment they are entered.

• Required fields: a customer cannot be saved without a tax number, address and payment terms.
• Formats: postcodes, phone numbers and bank details are checked against expected patterns.
• Pick-lists instead of free text for values such as country, industry or unit of measure.
• Duplicate checks that search for similar names and addresses before a new record is created.
• Approval steps for sensitive changes, such as supplier bank details.

Start with the rules your system already supports. Most business platforms include required fields, pick-lists and basic duplicate warnings that are simply switched off or never configured. Turning them on costs little and often removes a large share of new errors within weeks. More advanced matching, for example comparing addresses that are written differently, can follow once the basics are in place.

Limit who can create records. In many organisations, dozens of people can add a new customer or product. Concentrating creation in a small team, or routing requests through a simple form reviewed by that team, reduces errors sharply. It does add a short delay, so agree a response time that the business can live with.

Clean the backlog

Merge duplicates and correct errors in planned stages. Once new errors are being prevented, the existing backlog can be cleaned without it refilling behind you. Trying to clean everything at once usually fails, so work in priority order.

1. Measure the problem: count duplicates, incomplete records and inactive records in each data set.
2. Start where the business pain is greatest, such as active customers with open orders or high-volume products.
3. Archive records that have not been used for an agreed period, rather than cleaning them.
4. Merge duplicates carefully, keeping transaction history linked to the surviving record.
5. Check results with the people who use the data before moving to the next stage.

Consider a typical wholesaler that found nearly a fifth of its customer records were duplicates or inactive. Rather than tackling all of them, it started with the customers who had ordered in the past year, merged duplicates in weekly batches with sales and credit control checking each batch, and archived the rest. Within a few months, sales reports and credit exposure figures agreed for the first time in years.

Keep it clean

Make quality visible to the people who create the data. A short monthly note showing each team’s error and duplicate rates, without blame, tends to improve behaviour faster than any instruction. Celebrate improvements as well as pointing out problems.

Measures worth tracking

• Number of new duplicate records created in the month.
• Percentage of active records missing required fields.
• Average time from request to new record created.
• Number of reporting disputes traced back to master data.

Quality needs ongoing attention. Track a small number of measures monthly, such as new duplicates, incomplete records and time to create a new record, and share them with the owners. Review access rights periodically. When a new system or integration is introduced, check how it creates and updates master data before it goes live, because integrations are a common source of new duplicates.

Where to start

Be patient with the cultural side. People who have created records freely for years may see new rules and approvals as bureaucracy. Explain the cost of bad data in terms they recognise, such as delayed deliveries, disputed invoices or hours spent reconciling reports, and show them how quickly the new process works. Once they see fewer problems downstream, most become supporters.

Pick the master data set that causes the most disputes, usually customers or products. Name its owner, agree what a good record looks like and switch on the basic validation rules your system already supports. Then measure the backlog and plan the first clean-up stage. Fixing one data set properly will show the rest of the organisation what good looks like.

Source: Intelliversal analysis.

Frequently asked questions

Do we need a master data management system?

Not usually at first. Most business systems have enough validation and workflow features to make a large difference. Specialist tools become useful when many systems hold the same data and need to be kept in step.

Who should own customer master data?

Usually a business function close to its use, such as sales operations or credit control. IT supports with tools and access but should not decide what a correct record looks like.

How long does a master data clean-up take?

It depends on volume, but working in priority stages means the most important records can be cleaned within weeks, with the rest following over a few months.

Explore Data foundations and advanced analytics