Legacy System

A legacy system is software that is still doing important work in your business while nobody is actively developing it any more. It might be twenty years old or five. Age is not really the point.

The useful definition is not "old". It is: a system you depend on, that resists change, and that fewer people understand every year.

How to tell whether you have one

Not by the date it was installed. By these signs:

Changing it is frightening. Nobody is confident what else a change might break, so requests get refused or deferred rather than costed.

One person is the documentation. If a specific individual would have to be called on holiday, the knowledge lives in a person rather than in the business.

It cannot represent something you now do. The business has moved into subscriptions, or services, or a new territory, and the system was built for how things worked before. Workarounds have accumulated.

Nothing else talks to it. Information gets retyped out of it and into everything else. This is the most common and most expensive symptom.

It is out of support. The supplier is gone, or the version no longer receives updates. This is the one that is genuinely urgent when it appears.

What it actually costs

Rarely an invoice, which is exactly why legacy systems survive budget reviews that kill much cheaper things. The cost turns up as hours instead: rekeying, reconciliation, errors, questions that take half a day to answer, and the things you declined to do because the system would not support them.

For an established SME the rekeying line alone is typically between £4,000 and £25,000 a year, every year, and nobody ever approved it. It simply accumulated.

What to do about it

Not necessarily replace it. Replacing working software with newer working software is redecorating, and it carries all the risk of a real project with none of the return.

The test is not "is this old". It is "is this stopping us doing something we have decided to do?" If nothing follows from the change, leave it alone.

Where something does follow, the cheaper and more reversible move is usually to modernise around it: leave the system doing what it does well, stop people having to touch it directly, and connect it to things. A failed incremental step costs you weeks. A failed replacement can cost the trading year.

Replacement earns its place in four situations: the system is unsupportable, the data model genuinely cannot represent the business, integration would cost more than replacement, or the process is being redesigned anyway.