MSAMM
Back to course

Free preview · Implementing Manufacturing and Materials Management: From Empty Plant to Costed Work Order

Data quality before migration

The client's item data is worse than they think. Said bluntly rather than diplomatically, because the diplomatic version does not get acted on. This is true on essentially every manufacturing implementation, and it is better to say it in week two than to discover it in test cycle two.

The common defects. Duplicate parts under different numbers. Obsolete items nobody has flagged — often twenty to forty per cent of the master. That is not a typo: up to two items in five. Missing or wrong units of measure. Descriptions that are free text with no convention. Attributes held in spreadsheets outside the legacy system — which means the extract is not the data, and nobody will mention this until you ask. And parts referenced on BOMs that do not exist in the item master.

The assessment to run, in week two: counts by type, duplicate detection on description similarity, items with no transactions in two years, items missing critical attributes, and BOM-to-item referential integrity. Five queries, a day of work. In week two they are findings; in month four they are a crisis.

Cleanse in the legacy system or in staging. Never in the target. Loading dirty data and fixing it afterwards means fixing it twice — once to load it, and again in a live system where every correction is a change with an audit trail.

The obsolescence conversation. Not migrating thirty per cent of the item master is usually the right answer and always an uncomfortable one. Bring the numbers, not the opinion: "a third of these have not been transacted in two years" is a conversation; "we think a lot of this is dead" is an argument you lose.

And the last piece, which is about the sponsor rather than the data. Data cleansing is a client work package with a client owner. Set the expectation early. If nobody owns it, it does not happen — and the project slips for reasons that get blamed on the system. That last clause is why this matters to you personally: the data was always going to be bad, and the question is only whether it was a known work package with a name against it, or a surprise that became your problem.

That is the end of the free preview. The full course covers the rest of the curriculum, with the assessment and a certificate on completion.

See the full course