Ask an engineering team what they would do with a fifteen-year-old business application and most will say: rewrite it. Ask them to estimate, and the number will be optimistic by a factor most of us have learned the hard way.
The trouble with the full rewrite is not the engineering. It is that the business does not stop while you do it. Two years in, the new system is chasing a moving target, the old system still holds the real data, and nobody wants to be the person who signs off the switch.
Replace at the seams
The alternative is to route traffic through a thin layer in front of the old system, then move one capability at a time behind it. Reporting is usually the safest place to start — it is read-only, so a mistake is visible but not destructive.
Each slice ships to production on its own. The old system shrinks gradually. If funding is cut halfway through, you are left with a working hybrid rather than an abandoned rewrite, which is a materially better place to be.
Decide what is worth carrying forward
Legacy code accumulates rules nobody remembers agreeing to. Some encode genuine regulatory or contractual obligations. Some encode a workaround for a supplier who stopped trading in 2016.
Separating the two takes interviews, not just code reading. We have found the fastest route is to sit with whoever raises the most support tickets. They know which behaviour is deliberate and which everyone works around.
What the plan should look like
- Freeze new feature work on the legacy codebase, but keep fixing defects — a frozen system people still depend on breeds shadow spreadsheets.
- Put monitoring on the old system first, so you can prove the new slice behaves identically.
- Migrate data continuously rather than in one weekend cutover.
- Agree in advance what “done” means for each slice, in business terms.
The honest trade-off
Incremental modernisation takes longer in total and costs more in integration work. What you buy is the ability to stop at any point with something functional. For most organisations that optionality is worth more than the theoretical efficiency of starting fresh.
