Written by the team building the engine that runs behind the numbers.
Three carriers completed their first two quarterly closes on HyperCal. Average close time across the three participants fell from 16 days to 9 days. Here is what they said, and what we learned from running alongside them.
The signs are usually not dramatic. They accumulate: the close takes one day longer each year, the audit prep requires a separate week, the CFO asks for the final number before the model is finished.
Most actuarial reports are over-documented in areas regulators treat as routine and under-documented in the areas they actually scrutinize. Knowing the difference changes how you structure your next submission.
Smaller carriers face a recurring decision: maintain a full actuarial team, outsource to a consulting firm, or find a middle path. The answer changes when modeling infrastructure can be shared.
When pricing and reserving models run in separate silos with separate assumption sets, consistency checks happen manually at the end, which is too late. A different approach: run them from a shared assumption layer.
Automating the RBC calculation is straightforward for the formula parts. The risk sits in how your system handles assumption changes and whether the calculation trace is recoverable.
Software engineers solved the “which version ran in production last Friday” problem decades ago. Actuarial pricing teams are still answering that question with inbox search and file timestamps.
A regulator asking “show me how you got from raw triangle data to this reserve figure” is not a difficult question if your system logged every step. It is a very difficult question if it did not.
Based on conversations with actuarial and finance teams during our early-access program, the same three bottlenecks appear regardless of carrier size. None of them are modeling problems.
Regulatory review of an internal model turns largely on documentation quality, not model sophistication. This checklist covers the items examiners consistently ask for first.
The three standard IBNR methods each make different assumptions about historical patterns and expected loss ratios. Here is when each holds up and where each can mislead.
Most spreadsheet-based reserving setups fail the same way: the model that passed the prior year review is not the model that produced last quarter's numbers. A look at where version drift happens and what it costs.
