Reserving process problems rarely arrive as clear failures. The quarterly close produces a reserve figure that gets signed off and filed. The appointed actuary's opinion is delivered on schedule. From the outside, the process looks functional. The problems show up in the friction: the days spent on data preparation before the model can run, the senior actuary who cannot take leave during close week because the model only runs correctly when they run it, the reserve memo drafted in the last 48 hours before the filing deadline by whoever is available.
Teams living inside these processes frequently normalize the friction because each individual instance of it is manageable. What is harder to see is the cumulative cost: senior actuarial time absorbed by tasks that should not require expert judgment, documentation that is just adequate enough to file but not adequate enough to support a thorough examination, and the systematic underinvestment in the analysis that genuinely requires actuarial expertise.
The five patterns below describe reserving processes that are functional but not sustainable. They do not require catastrophic failure to be worth addressing; they represent infrastructure problems that compound over time.
Sign One: The Close Can Only Run When One Specific Person Is Available
If your quarterly close requires one specific actuary to be present and available throughout the process, because they are the only person who fully understands the model structure, knows which cells link to which source data, or can interpret the outputs correctly, your reserving process has a key-person dependency that is a risk in its own right, separate from the infrastructure risk.
This pattern is not unusual in small actuarial teams where one person built the model and has maintained it for several years. The model may work well; the problem is that the knowledge required to run it correctly and interpret its outputs exists only in one person's head, and the model's documentation is insufficient to substitute for that knowledge. When the key person is on leave, or leaves the organization, or is unavailable during a close due to illness, the team either delays the close or runs the model with reduced confidence in the outputs.
The test for this sign: ask whether someone new to the team could run the quarterly close correctly using only the documentation in your model and process files, without asking the key person for guidance. If the answer is no, the process has a single-point-of-failure that is a documentation problem masquerading as a capability problem.
Sign Two: Answering "What Changed from Last Quarter?" Takes More Than an Hour
A question that management, auditors, or regulators ask routinely is what drove the change in the reserve from the prior quarter. The answer has two parts: what changed in the underlying loss data, and what changed in the assumptions or methodology applied to that data. A reserving process with good infrastructure can answer this question in minutes by comparing the current quarter's model run to the prior quarter's archived run and showing which inputs and assumptions changed.
A reserving process without good version control answers this question by having the actuary reconstruct it from memory and from whatever documentation was written at the time of the prior close. If the prior quarter's model was updated in place rather than saved as an archived version, the reconstruction requires recreating the prior quarter's calculation from raw data, which is time-consuming and introduces uncertainty about whether the recreation exactly matches what was originally filed.
This sign is particularly visible when it is needed most: during an examination, a management presentation, or an audit. The team that can pull up a side-by-side comparison of the current and prior quarter's assumptions and show exactly where the development factors, tail factors, and method selections changed is demonstrating model control. The team that must reconstruct the comparison from memory and archived email threads is demonstrating the absence of it.
Sign Three: Your Reserve Documentation Is Written After the Decision Is Made
Reserve documentation written at the time actuarial judgment is exercised is qualitatively different from documentation written afterward to describe decisions that were already made. Documentation written contemporaneously captures why the actuary chose a particular development factor over alternatives that were also considered, what data anomalies were observed and how they were handled, and what the uncertainty around the selected value is. Documentation written after the fact describes what was done, but the "why" has already degraded.
The pattern that indicates this problem: the reserve memo is typically the last document completed in the close cycle, drafted in the final days before the deadline, after the reserve figures are confirmed, by whoever has time to write it. The memo is often drafted by someone other than the actuary who made the key judgment calls, based on a description of the methodology rather than direct knowledge of the selection process.
This documentation approach is adequate for most routine examinations, where the reserve figures are accepted without detailed review of the supporting rationale. It is inadequate for targeted examinations focused on adverse development, audit reviews under IFRS 17, or rate hearings where the regulator asks how the reserve development assumptions connect to the filed pricing assumptions. In those contexts, documentation written after the fact is less credible than contemporaneous documentation and requires supplemental explanation that the contemporaneous documentation would have made unnecessary.
Sign Four: Your Close Window Has Not Shortened Despite Your Team Growing
In a well-functioning reserving process, adding an actuary to the team should compress the close window, or maintain the window while adding analysis capacity. If your team has grown over the past two or three years but the close window has stayed the same or lengthened, the bottlenecks are in the infrastructure rather than in the headcount. The additional actuary time is being absorbed by the same data preparation, sequential review dependencies, and documentation assembly that consumed the prior actuary's time.
This is a diagnostic sign rather than a definitive problem, because some close window length is appropriate and reflects the time needed for thorough analysis. But a close window that is stable regardless of team size, or that grows proportionally with team size rather than shrinking, indicates that the process has structural bottlenecks that additional people cannot solve. The bottlenecks are typically in the data assembly phase, which is time-sequential regardless of how many people are available, and in the review chain, which becomes longer rather than shorter as the team adds senior actuaries who each require a review step.
Sign Five: You Cannot Recover the Calculation for a Specific Prior Quarter Without Significant Effort
This sign is the most direct indicator of an infrastructure problem. If a regulator, auditor, or internal review requested the complete calculation workpapers for a specific quarterly close from two years ago, including the input data, the development factor selections, the method choices, and the final reserve by segment, how long would it take your team to produce those materials?
In a process with proper version control and archived calculation artifacts, this is a data retrieval operation: locate the archived run for the requested quarter, export the relevant exhibits, and package them for delivery. In a process where quarterly model runs updated the working model in place, where input data was not archived alongside the model, and where documentation was not linked to specific model versions, the response requires reconstruction rather than retrieval. Reconstruction is time-consuming, may be incomplete due to subsequent updates to source systems, and cannot be verified to exactly match the original calculation.
For carriers subject to NAIC financial examinations, state departments typically request actuarial workpapers as part of the standard examination package. For carriers subject to IFRS 17 audits, the external auditor requires contemporaneous documentation for the current period and may request documentation from prior periods for comparison. In both contexts, the inability to quickly produce a complete and verifiable record of a prior period's calculation is a documentation deficiency that the team needs to explain, even if the reserve figures themselves were appropriate.
The Underlying Pattern: Infrastructure Debt Accumulates Slowly
The five signs above are not independent problems with independent solutions. They are different surface expressions of the same underlying condition: the reserving infrastructure was built to produce a quarterly output, not to support the review, documentation, and knowledge transfer requirements that accumulate around that output over time. A spreadsheet-based model that runs correctly and produces an appropriate reserve can still score on all five signs if it was built without attention to version control, documentation, and operational resilience.
Addressing these signs does not mean replacing the actuarial methodology. Chain Ladder, Bornhuetter-Ferguson, and Cape Cod methods that have been calibrated to a carrier's book over years are not the problem. The infrastructure that runs those methods, archives the results, and maintains the documentation record is where the modernization investment needs to go. The goal is not a different actuarial analysis; it is an infrastructure that makes the analysis transparent, recoverable, and independent of any one person's institutional memory.
We recognize that these infrastructure investments compete with other priorities in a small to mid-size carrier's budget. The argument for making them is not that the current infrastructure is failing; it is that the risks it creates compound over time, and the costs of addressing them before a regulatory examination or an audit request are substantially lower than the costs of addressing them under time pressure afterward.
