Test Case Study — Six days to four hours on a claims platform
Test fixture. A 12-year-old claims monolith priced a policy change in six days. Eleven weeks later it took an afternoon.
- Event-driven architecture
- Strangler-fig migration
- Regulated
The story
How we approached the work, what we built, and why it matters.
The problem
Pricing logic lived in 40,000 lines of stored procedures that three people understood and none of them fully. A rate change meant a six-day cycle of manual regression testing, and the compliance team could not get a straight answer about which version had priced any given policy.
The approach
We put an anti-corruption layer in front of the pricing path, lifted the rules into an event-sourced service, and shadow-ran it against live production traffic for six weeks — comparing every output against the legacy engine before a single customer saw the new path.
Disagreements between the two engines were treated as findings, not bugs to be smoothed over. Eleven of them turned out to be latent defects in the original system.
The outcome
Rate changes now ship in an afternoon. Every priced policy carries the exact rule version that produced it, so the compliance answer is a lookup. The client's own engineers have shipped four rule changes since we left, without us.
Related case studies.
All case studiesHave a similar problem worth solving?
Tell us where you’re headed. We’ll bring the engineering discipline to get you there.