A completed-case median can be a useful statistic. The trouble begins when the words completed cases disappear and the same number is presented as if it describes every claim in the cohort. The cases still open have not supplied completion durations. Their absence from the calculation is therefore part of the meaning of the result, not a technical detail that can be safely removed from the management account.
The fictional Soravane example concerns 11600 claims in a 2025 cohort. At the reporting cutoff, 9500 were completed and 2100 remained open. The completed cases had a median duration of eighteen days. I retained that explicitly labelled statistic and introduced the count and age distribution of the open group beside it, through JPY 270 million of reporting reconstruction. I did not produce an all-cohort median, a forecast of eventual completion or a reserve conclusion from those observations.
Keep the population attached to the median
I first want to know how the cohort and the observation cutoff were defined. Without those conditions, the number can appear comparable to another report even when the populations differ. A measure calculated from cases completed during a period is also not necessarily the same as a measure for a starting cohort observed until a cutoff. The question should be specified before the analyst chooses whichever population produces the most familiar dashboard.
In this example, the eighteen-day median says something about the 9500 cases with observed completions. It does not tell us the eventual completion duration of the 2100 still open. The open cases might include matters with different circumstances or ages. We need to inspect that population rather than assume its eventual outcomes will reproduce the completed group's distribution. The arithmetic of the median can be correct while the broader sentence used to describe it is unsupported.
The reporting sponsor wanted one number that could travel easily between meetings. I understood that preference. A long methodological discussion is not a useful substitute for a clear operating account. My objection was to making the number easier to carry by removing the condition that gave it meaning. A concise label and an adjacent view of open cases offered a way to retain practical usability without suggesting the completed subset represented an outcome already observed for everyone.
I also wanted to avoid treating a change in the completed-case median as a complete explanation of service performance. The mix of cases completing by a cutoff can change. A different median may prompt useful questions, but it does not by itself identify the cause or establish how every claim is progressing. Claims management owns that operating discussion. The data function helps by making the population and the observation conditions clear enough for the discussion to begin on a defensible basis.
Do not turn present age into finished duration
One tempting response is to include the open cases by treating their ages at the cutoff as if they were completion times. That gives every case a number, but it changes the meaning of the measurement. An open claim aged twenty days has been observed for twenty days without a completion. It has not been shown to complete in twenty days. Combining that age with a finished duration can create an apparently inclusive statistic that no longer answers the intended question.
I prefer to make the two kinds of observation visible. The completed group can show its observed duration account. The open group can show how long those cases have been open at the cutoff, with definitions suitable for the operational question. The example does not prescribe a universal dashboard or a particular statistical method. It establishes the more basic requirement that present age and completed duration should not be made interchangeable simply to fit one headline.
There is a legitimate objection that two views can make comparison harder. A senior reader may ask how to judge progress without a single overall measure. I would respond by clarifying the decision they need to make. If the question concerns current workload, the open population matters directly. If it concerns completed service, the completed account remains useful with its conditions. If it requires a forecast, that requires an explicit forecasting approach and evidence beyond pretending that incomplete observations are finished ones.
My earlier closure-reporting mistake makes me cautious about the language here. In 2007 I described closure events as final resolutions, and 260 of the associated claims reopened within ninety days. I had attached certainty to an event that did not establish it. The duration problem is different, but it poses a related leadership test: can I keep a convenient operational label from becoming a stronger statement than the evidence supports? The answer has to remain open to challenge by colleagues.
Let the report support a more precise conversation
The reporting reconstruction gave management a clearer view of observed completion and continuing work. It did not itself shorten claim handling or prove that a new allocation decision would succeed. I want that distinction retained in a public case study because data leaders can easily claim the whole operating result that follows a report. Here, the documented output was the account supplied to the decision-maker. Any subsequent service change would need its own evidence and attribution.
Actuarial and finance colleagues also have questions that may involve claims information, but an operational duration report does not settle those questions by proximity. I would not infer a reserve position from the eighteen-day median or from the current ages of the open group. The relevant professional owners need information and methods appropriate to their responsibilities. Keeping that boundary visible helps the CDO explain a substantive contribution without presenting data leadership as authority over every use of the data.
For a board reader, I would make three conditions easy to find: the cohort being followed, the cutoff at which its status was observed and the population contributing to the stated median. An account of ages of the open group should be intelligible alongside them. The board can then ask management what the combined picture means for the decision at hand. It does not need to reconstruct every calculation, but it should be able to see where the observation is unfinished.
All people, figures and events in this essay are invented. A real account would require verified definitions and permission to describe the material without revealing policyholder information. My argument is that the unfinished population should remain part of the story precisely because its outcomes are not yet known. An eighteen-day median can be worth reporting. It becomes more useful when the reader can see whose completed experience it describes and which claims are still waiting to supply the rest of the evidence.