How long does custom software last?

Nothing rots, but everything around it moves

The code you paid for will do exactly the same thing in ten years as it does today. That is the part people mean when they ask whether software expires, and the answer is no.

The reason it still needs attention is that it does not run in isolation. It sits on platforms that update, talks to services that change, handles data under rules that get rewritten, and serves a business that does not stand still. Every one of those moves underneath it.

The four things that actually age

The foundations it runs on. The building blocks underneath any modern system get security updates, and old versions eventually stop getting them. Staying roughly current is routine and cheap. Skipping it for three years turns a series of small updates into one large, risky job.

The services it connects to. Payments, email, maps, accounting, anything it talks to can change how it works or retire an old way of doing things, usually with notice. Something has to be watching for that notice.

The law. Data protection, accessibility, payments and consumer rules all move. A system that was compliant when it launched is not automatically compliant five years later, and this is the one that carries a real penalty rather than an inconvenience.

The business. This is the fastest of the four. What you sell, how you price it and who does what change constantly, and software built around the old shape gets worked around until the workarounds are the system.

What looking after it involves

In practice it is unglamorous and mostly invisible: keeping the foundations current, watching for breakages before anyone reports them, keeping backups that have actually been restored at least once, and making the small changes the business asks for.

This is what a maintenance arrangement should be buying, and it is worth asking exactly what is included. Hosting on its own is not maintenance. Neither is a promise to fix things when you notice them, because by then a customer has noticed too.

The signs it is coming to the end

The usual first sign is that changes stop being cheap. Something small takes a week and nobody can quite say why, because each change has to be threaded through years of things nobody wants to disturb.

The second is that the workarounds have become the process. People keep a spreadsheet beside the system, or use a field for something it was never meant for, and training a new person now means teaching them both the software and the ways around it.

The third is that the business it was built for no longer exists in that shape. That is not a failure of the software, it is what success looks like, and it is the most common honest reason to rebuild.

Rebuild or repair

Repair is right far more often than people expect. If the shape is still correct and the trouble is in one area, replacing that area is a fraction of the cost and carries a fraction of the risk.

Rebuild when the shape itself is wrong, because everything you add to the wrong shape makes the eventual move harder. A rebuild also tends to be quicker than the original: the hard part of the first build was deciding what it should do, and that decision is already made and proven.

Related questions