Mining the hidden gold in eight years of legacy code

When an eight-year-old legacy application has to be replaced, the common assumption is that it should be thrown away as fast as possible. We made the opposite choice: treating that legacy as a gold mine rather than a burden, because the fine truth of the calculation, its precise rules and its exact mechanics, had crystallised in the code over the years.

The business had a perfect grasp of the concepts, but not of the detail nor the exact method used at the time, a level of precision that was nonetheless essential in a legal framework requiring the same calculation method to be reproduced faithfully.

The context: a critical calculation application in the Belgian public sector, handling legal data, where error was simply not an option, and where the constraint was non-negotiable: a big-bang go-live. No way, then, to follow a classic analyse-develop-deliver iterative cycle.

We extracted the hidden gold from an existing battery of business tests and rebuilt the whole back end in Test Driven Development from that raw material, with business analysis supporting the technical analysis rather than preceding it. I'll share how this meticulous approach allowed us to reach one hundred percent conformity with the existing system, then how a Dual Run deployment, both systems running in parallel with automatic comparative reporting, let us secure every discrepancy in real conditions on production data before the final switchover.