During the early-access program we ran with a small cohort of non-life carriers over the first half of 2025, we spent time before each close cycle asking actuarial and finance teams to walk us through their process in detail. Not the idealized version in their procedure documentation, but the actual sequence of events for the most recent close, including the parts where things slowed down.
The bottlenecks that emerged from those conversations were consistent enough across carriers of different sizes, different lines of business, and different operating environments that we concluded they represent structural properties of the spreadsheet-based quarterly close workflow rather than problems specific to any one team.
Three bottlenecks appeared in every case. None of them were modeling problems. None were symptoms of an underskilled actuarial team. They were all workflow and coordination problems that a different kind of infrastructure would prevent.
Bottleneck One: Data Assembly Before the Model Can Run
The quarterly close typically cannot begin until the actuarial team has assembled a complete and reconciled data set: claims triangles by line of business, premium registers, paid loss data, case reserves, a rate change index for pricing models, and sometimes a separate reinsurance treaty position report. Each of these typically comes from a different source system, arrives at a different time, and requires manual reconciliation with the prior quarter's version.
In conversations with participating teams, the data assembly phase consumed between two and four days of the typical three-week close window. The time was spent not on analysis but on extraction, format normalization, identifying discrepancies between sources, chasing the IT or finance contacts responsible for the source systems, and re-running extracts after corrections.
This phase is entirely upstream of the actuarial model. No reserve judgment is being made. No assumption is being reviewed. The actuaries are functioning as data translators, bridging source systems that were not designed to talk to each other and that produce outputs in formats requiring manual transformation before they can enter the reserving model.
The pattern holds across carriers that use modern core systems and carriers still running older administration platforms. The bottleneck is not the source system quality; it is the absence of a standardized, automated handoff from source systems to the actuarial calculation environment. When that handoff is manual, it adds days to the close window regardless of how good the downstream model is.
Bottleneck Two: Sequential Review Under Time Pressure
Once the model runs and produces initial outputs, the results move through a review sequence: the actuary who ran the calculation reviews it, then it goes to a senior actuary or appointed actuary for substantive review, then to the CFO or finance director for sign-off, and in some cases to the board risk committee before the figures are finalized. These reviews are intended to be independent quality checks, but in practice they become sequential dependencies that compress the calendar.
The senior reviewer cannot begin until the initial outputs are available, which is determined by the data assembly timeline. The finance sign-off cannot begin until the senior review is complete. At a carrier where the quarterly close window is three weeks, the data assembly phase consuming four days and the sequential review chain adding another four to six days of calendar time means the final figures are being confirmed with less than two weeks remaining before the statutory deadline.
When something unexpected appears in the initial outputs, this gets harder. An IBNR that is materially higher than prior quarter for a line that has not had obvious adverse development triggers a question: is this a real change in the loss environment, a data issue, or a model artifact? Answering that question thoroughly requires pulling together the claims data, the prior quarter model run, and the trend analysis, and that investigation pushes the review timeline out further.
The underlying problem is not that review takes time. Review should take time; it is the mechanism that catches errors and exercises appropriate oversight. The problem is that review happens sequentially after outputs are produced rather than being structured as a concurrent process where the reviewer can monitor calculation progress and begin interpreting results before the full output is available. A platform that provides reviewers with real-time access to intermediate outputs and a shared log of model changes in progress changes the review dynamic from a queue into a parallel process.
Bottleneck Three: Documentation Assembly Running in Parallel
The third bottleneck is the one teams talk about least, possibly because it feels like overhead rather than a core function. Documentation for the quarterly close includes the reserve memo, the assumption change log, the basis of preparation for the financial statements, and any supporting materials required for the appointed actuary's report. For carriers with regulatory reporting obligations, add the relevant actuarial sections of the statutory filing.
In most teams we spoke with, this documentation is drafted in parallel with the close process and finalized after the reserve figures are confirmed. Because the documentation references specific model outputs, it cannot be completed until those outputs are final. Because the close timeline is compressed, the documentation phase occupies the last few days of the window, which is also when the signed-off figures need to be entered into the financial reporting system.
The result is that documentation is consistently under time pressure, tends to be drafted by whoever is available rather than by the person who made the underlying judgment calls, and often describes what was done rather than why. The reserve memo that says "development factors were selected consistent with the three-year weighted average, excluding the anomalous 2022 development period" is a description. The memo that says "we excluded 2022 because the company changed its claims management process in Q3 2021, accelerating closings; including that period would systematically understate IBNR for current accident years" is documentation. The distinction matters when a regulator asks six months later.
Documentation generated contemporaneously with the model run, capturing the actual rationale as it is being formed rather than reconstructing it afterward, addresses this bottleneck structurally rather than through documentation process instructions that teams do not have time to follow.
What These Bottlenecks Have in Common
All three bottlenecks share a root cause: the infrastructure supporting the quarterly close was not designed with the close cycle's specific coordination requirements in mind. Spreadsheets are calculation tools; they do not manage data handoffs, support concurrent review access, or capture rationale as a structured record attached to the calculation. Building workarounds to spreadsheet limitations is where most of the close cycle's elapsed time goes.
We want to be direct about what we are not saying here. We are not saying the teams we worked with were inefficient. They were experienced, skilled actuaries working within the constraints of tools that were not designed for this purpose. The inefficiency is in the infrastructure, not in the people. When we built HyperCal, the three bottlenecks described here were the primary design targets. Not because we identified them from first principles, but because the teams who walked us through their actual close process told us exactly where the time was going and what a different infrastructure would need to do differently to free it up.
