Aripone / case study
From 27 hours to 8 per run.
Finding and fixing the slow path in an insurance-sector calculation engine. Results delivered by our principal in regulated production before Aripone was formed.
- Client
- An insurance-sector company
- Team
- Our principal, Rey Narvaez
- Outcome
- 27 h to 8 h per run
- Date
- Early 2026
Figures are run time per run of the engine, measured before and after the change.
The problem
The company ran a legacy calculation engine in a regulated environment. Each run took 27 hours, and the job ran back to back on weekdays. A large downstream run, which compares results across all insurance policies and has to be accurate, could not start until the engine finished.
Part of the cost was in how the engine talked to its dependencies. On every operation it opened and closed its third-party connections, cloud storage among them, and each open and close cost compute and wall time. Its database queries were the other part.
The fix
- The database and cloud connections became singletons, opened once and reused.
- Record processing was restructured to run in parallel.
- Queries got fixed indexes, lost their unnecessary joins and cartesian products, and their results are now handled differently in code.
The outcome
- before27 h
- after8 h
8 hours per run, about 70% shorter. That is 19 hours less run time per run. Three runs now fit in the time one used to take, so the job could run more often.
The client's insurance-policy domain expert checked the new output by hand against the base run results, and the policy-wide comparison started sooner.
Your slow path
Find your slow path in five days.
The Performance Diagnosis is five working days at a fixed fee, published on the pricing page. On day one we reproduce the slow job, pipeline or hot path outside production and agree the baseline metric. On day five you get findings ranked by time or cost recovered, a roadmap ordered by effort, a benchmark plan to prove the fix, and a readout. We quote the fix as a separate fixed-price SOW, and Aripone names the engineer who leads the diagnosis on the SOW.