Data Quality

The same customer, four different ways

Master data clean-up is reconciling your customer, supplier, item and site records to one version — and then putting standing checks in place so the mess does not quietly rebuild itself over the next three years.

Symptoms

If people don’t trust the data, they stop using it

They start keeping their own version instead, which is how one business ends up with four answers to the same question and a month-end reconciliation nobody can explain.

Two reports on the same question disagree

The same supplier appears four times

Three sites classify a product three ways

People keep their own spreadsheet instead

Month end includes a manual reconciliation

A migration slipped because of the data

An AI pilot returned answers nobody believed

What gets reconciled

Four masters. Everything else hangs off them

Almost every reporting argument in an operating business traces back to one of these four being inconsistent.

01

Customer master

One customer, one record, one credit limit — with the old spellings kept as aliases so nobody loses their history.

02

Supplier master

Duplicates merged, ABNs verified, payment terms reconciled to what was actually agreed.

03

Item and product master

One classification across every site, so a group number means the same thing everywhere.

04

Site and location master

Consistent addresses, codes and hierarchies, which is what makes site comparison possible at all.

Keeping it clean

A clean-up without standing checks is a temporary result

Master data degrades because people are doing their jobs quickly, not because they are careless. The answer is checks that catch drift while it is one record rather than four hundred.

Duplicate detection on new records
Format and completeness checks
Classification consistency across sites
Standing reports on what changed
Exceptions raised while they are still small
Rules owned and adjustable by your team

Why this comes before AI

An AI assistant answering from inconsistent master data will give you a confident wrong answer. This is the work that has to happen first, and it is why HeyMi and Data Mi are better together even though neither needs the other.

Your data

Handled carefully

Your data is processed in Australia
Working data is held in Mattingly’s private cloud environment
Transfers use encrypted methods including HTTPS and SFTP
You control how long anything is retained
Your data is not sent to external AI services
Source and intermediate files are deleted on schedule or on request

What you get at the end

Reconciled masters, the rules that produced them, the exception list, and the standing checks — documented, and running in your environment.

Questions

Data quality, answered

Data quality is whether the information in your systems is accurate, consistent, complete and trusted enough for people to act on. In practice the symptoms are recognisable: two reports disagree, the same customer exists four times, and people keep private spreadsheets because they do not believe the system.

By profiling the data and matching on the attributes that actually identify a business — name, ABN, address, contact details — with rules your people agree to. Likely matches are merged automatically. Uncertain matches are escalated to somebody in your business, because a wrong merge is worse than a duplicate.

Because AI amplifies it. A report with duplicate customers produces a number a person can sense-check. An AI answer built on the same data produces a confident sentence with no visible working. If people do not trust the data, they stop using what is built on it.

No. Data quality work is worth doing where a decision depends on it. Fixing the customer master because pricing decisions run through it is worth the effort. Fixing a field nobody reads is not.

Talk to us

How many ways does your biggest customer appear?

Most people guess two. It is usually four or five once you count the systems nobody mentions.