Translate rating rules side-by-side with actuaries instead of waiting for a consolidated spec
Context
A US personal-lines carrier had committed to retiring a PL/I + IMS-DB rating engine, but no one could produce a complete list of the rules it executed. They lived as 100+ hardcoded procedures over IMS-DB lookups, authored over two decades by actuaries who had since rotated out. The obvious move — request a spec, wait for one, escalate when it doesn't arrive — would have killed the project on the same rock every mainframe re-platform dies on. The rules system and execution system were fused at runtime, which is the actual reason re-platforms fail.
Decision
Stop asking for the spec and produce it, one rule at a time, in pairs with whichever actuary owned that line of business. Sit next to the SME, read the PL/I together, express the same decision tree in Ratabase semantics, and run both engines side by side on a curated policy set until they reconciled cent-for-cent. Each rule gets signed off by the actuary and switched independently — Ratabase becomes the source of truth for that rule, while the mainframe keeps executing the others nearby.
Consequences
- 100+ rating algorithms translated and switched without a write freeze.
- The rule set lived half on each engine for years — fine, because the contract (given these inputs, return this premium) never moved.
- Individual re-rate cycles triggered by reconciliation errors returned several-hundred-thousand-dollar customer refunds — proof the new engine was load-bearing on real money.
- Established the pattern that ambiguity at this scale isn't waiting for a missing artifact, it's the absence of anyone whose job is to produce it — you either appoint yourself or you wait forever.