Legacy modernisation: replace old systems without stopping operations

Big-bang rewrites almost always fail. How to renew a grown system step by step — while it stays in production.

The old system is slow, expensive to maintain and nobody dares touch it anymore. The “rewrite it all at once” reflex is understandable — and usually a mistake. Big-bang rewrites take longer than planned, and by the time they ship, the old system has already absorbed new requirements.

Step by step, not big bang

The proven pattern is the strangler fig: the new grows around the old until the old is redundant.

  • Put a facade in front of the legacy system that routes requests.
  • Carve out one function at a time into new, well-tested services.
  • Keep the data flow in sync until the old path can be switched off safely.

Each step delivers value on its own — and can be rolled back individually. That cuts the risk dramatically.

The real win

Modernisation is not an end in itself. The goal is a system that is changeable again: faster to evolve, cheaper to run, understandable for new team members. We order the steps so the biggest pain point disappears first.

← All articles