Legacy software earns a bad name it does not always deserve. A fifteen-year-old system that still matches how the business works, can be hosted safely and can be changed by someone who understands it is not a crisis. It is an asset.
The problem starts when the software can no longer absorb change. New products take months because the data model will not stretch. The only person who can deploy it is a contractor who last visited in 2019. A security update would require a rewrite. Staff have invented shadow processes around the gaps.
Replacement is expensive. So is staying. The useful test is not age. It is optionality: can you still change the thing, recover it, and hire someone to look after it without a treasure map?
Sometimes the answer is a staged rebuild. Sometimes it is a new platform beside the old one. Sometimes it is leaving the core alone and replacing the edges. The wrong answer is a big-bang rewrite sold on the promise that everything will be nicer afterwards.