The short answer
Claims operations have a distribution problem that most reporting hides.
The majority of claims close quickly and consume little effort. A minority run long, span multiple systems, require repeated medical or technical assessment, and stay open for months or years. That minority consumes a disproportionate share of handling capacity.
Average handling time, the metric most claims operations report, conceals this completely. A distribution with most cases at days and a tail at months produces an average that describes neither group.
The effort in the tail concentrates in three activities that appear in no process documentation: assembling the status of a case across several systems, requesting updates from parties who owe information, and documenting the reasoning behind decisions for audit purposes.
Those three scale with case age rather than with case volume, which is why a stable claim count can coincide with a growing operations workload.
Key takeaways
- Average handling time hides a bimodal distribution, and the tail carries the capacity.
- Status assembly across systems is the largest single manual cost in long-running claims.
- Chasing information from external parties is coordination work that produces no progress on its own.
- Documentation for audit is required and frequently unsupported by the systems it documents.
- Effort in long-tail claims scales with case age, so workload grows without claim volume growing.
- Regulatory obligation and system compensation look identical, and separating them is where the recoverable hours are.
Why long-tail claims consume disproportionate effort
Status has to be assembled
A claim that has been open for eight months has touched a core policy system, a claims platform, a medical or technical assessment system, a payments system, and probably a spreadsheet or a shared document.
No single system holds the current state. Establishing where the case stands requires a person to check several places and reconcile what they find, and that assembly happens every time the case is reviewed, escalated, queried or reported on.
The cost is per review rather than per claim, which is why it grows with case age.
Progress depends on external parties
Long-running claims wait on information from outside the organization: a medical report, a technical assessment, a regulatory response, documentation from the claimant.
The handler cannot advance the case and can only follow up. That follow-up is real work, it repeats on a cycle, and it produces nothing except the possibility that the information arrives.
Most claims systems track case status. Few track who owes what and when it was last requested, which means the follow-up schedule lives with the handler.
Decisions require documented reasoning
Every non-standard decision on a claim has to be recorded with its rationale, for audit and regulatory purposes. That obligation is legitimate and it is frequently unsupported by the system, which captures the decision and not the reasoning.
The result is documentation written in free text, in a system field or a separate document, which is both effortful to produce and difficult to query afterward.
Handover loses context
Cases running for months outlast team assignments. When a case transfers, the receiving handler reconstructs the history, which means reading through the assembled record and calling whoever handled it before.
The reconstruction cost is proportional to how long the case has run and how much of its history lives outside the system.
Where the effort concentrates
| Activity | Typical driver | Recorded anywhere? |
|---|---|---|
| Status assembly across systems | No single view of the case | No |
| Follow-up on external information | Nothing tracks who owes what | Partially, in handler notes |
| Decision documentation | System captures decisions, not reasoning | In free text, hard to query |
| Handover reconstruction | Case outlasts team assignment | No |
| Exception routing | Non-standard cases have no defined path | No |
| Regulatory reporting assembly | Data spread across systems | No |
Every row is work that a claims operation performs continuously and that no claims metric measures.
The metric problem
Claims operations report volume, average handling time, closure rate, leakage and customer satisfaction. All of these are useful and none of them reveal the distribution.
Three changes make the reporting informative.
Report handling time as a distribution, not an average. The median and the ninetieth percentile together say more than the mean, and the gap between them is the size of the tail.
Report effort per case band. Hours consumed by cases closing within a week, within a month, and beyond three months. In most operations the third band is a small share of cases and a large share of hours.
Report wait time separately. The share of a case's elapsed duration spent waiting for an external party, which is the portion no internal process improvement will address and the portion that follow-up mechanisms can reduce.
Claims effort distribution
Where the capacity goes
| Band | Share of cases | Share of effort | Dominant activity |
|---|---|---|---|
| Closes within days | Large | Small | Standard processing |
| Closes within weeks | Moderate | Moderate | Assessment and validation |
| Runs for months | Small | Large | Status assembly, follow-up, documentation |
| Disputed or litigated | Very small | Disproportionate | Reconstruction and evidence assembly |
Average handling time reports a number that describes none of the four rows. The operational question is always about the bottom two.
The required versus compensating distinction
Manual work in claims falls into three categories that are indistinguishable in a process map.
Required by regulation. Assessment by a qualified person, documented reasoning for a decision, retention of evidence. These can be supported and cannot be removed.
Required by consequence. A payment decision that cannot be reversed, a communication to a claimant that cannot be unsent. A person decides because the error is asymmetric.
Created by system limitation. Status assembly across five systems, manual follow-up scheduling, reconstruction on handover, reporting built by hand. None of this is required by anything except the current configuration of the estate.
The third category is where the recoverable hours are and it is frequently described as belonging to the first, by people who inherited the step rather than designed it. Separating them requires asking why each step exists and whether the reason still holds, which is a discovery question rather than a documentation one.
What to establish before redesigning
- The handling time distribution, with median and ninetieth percentile
- Effort per case band, in hours
- For long-running cases, how many systems hold part of the state
- How much time per case per review goes into status assembly
- What proportion of elapsed time is waiting on external parties
- Which decisions require documented reasoning and how that documentation is produced
- How often long-running cases change handler, and what reconstruction costs
- For each manual step, whether it is regulatory, consequence-driven or compensation
Items 4 and 8 are the two that most often go unmeasured and the two that determine where the redesign should start.
Where Horizon fits
Horizon is an AI-powered continuous discovery platform. In claims operations its role is quantifying the effort in the long tail and separating required work from compensation.
Discovery Cycles run AI-led interviews across handlers, assessors and the functions they depend on, adapting to each role and following up on why a step exists rather than only recording that it does. The Insights Dashboard ranks findings by effort and impact with traceability to the input behind each one. The Process Library structures the resulting documentation, which is what makes a process map usable for redesign rather than for filing.
La Segunda Group, one of Argentina's main insurance companies with more than 90 years of experience and 1,200 offices nationwide, ran exactly this analysis on its long-term injury claim process.
The process was long and manual. Teams used several systems, including three named platforms plus spreadsheets, with little integration. More than 700 active cases were handled with high effort and limited visibility, which is the status assembly problem at scale. The organization wanted to simplify the process, reduce manual work and gain transparency, and needed to find friction points and repetitive tasks before designing improvements.
Horizon interviewed seven case managers and two medical auditors, covering 100% of the team involved in the process. The full process was mapped and analyzed in 48 hours. In under five weeks the engagement delivered a complete BPMN process map, a dashboard of insights grouped by effort and impact, and 10 key findings covering automation, integration and workflow redesign opportunities, alongside improvement initiatives for traceability and efficiency. It saved more than 30 discovery hours against the manual approach.
Conversation satisfaction was 8.2 out of 10 and 90% of participants said they would talk again, which matters in a regulated operation where handlers describe their own workarounds only when the exercise does not feel like an audit.
The results were used to begin building a new medical follow-up platform for chronic case management, with the dashboards guiding next steps on automation, alerts and system integrations. Chronic case management is precisely the long-tail band described above, which is where the analysis pointed.
That is one engagement under specific conditions rather than a projection for any organization.
Claims operations checklist
- Do you report handling time as a distribution rather than an average?
- Do you know effort per case band, in hours?
- How many systems hold part of the state of a long-running case?
- How long does status assembly take per case, per review?
- What share of elapsed time is spent waiting on external parties?
- Is there a mechanism tracking who owes what and when it was last requested?
- How is decision reasoning captured, and can it be queried afterward?
- How often do long-running cases change handler, and what does reconstruction cost?
- For each manual step, is it regulatory, consequence-driven or compensation?
- Do handlers and assessors contribute when the process is reviewed, or only process owners?
FAQ
Why do insurance claims operations struggle with long-running cases?
Because effort in those cases scales with age rather than with volume. Status has to be assembled across several systems every time the case is reviewed, progress depends on external parties who have to be chased, decisions require documented reasoning the system often does not support, and cases that outlast team assignments require reconstruction on handover.
What is wrong with average handling time as a claims metric?
Claims handling time is typically bimodal: most cases close quickly and a minority run for months. An average describes neither group, and the minority carries a disproportionate share of the effort. Reporting the median alongside the ninetieth percentile, and effort per case band, gives a usable picture.
Where does manual effort concentrate in claims processing?
In status assembly across systems, follow-up on information owed by external parties, documentation of decision reasoning, and reconstruction when a long-running case changes handler. None of these appear as steps in process documentation and none produce a record in the claims system.
Can claims processes be automated in a regulated environment?
The analysis can, generally. The decision, conditionally. Steps where a wrong output produces irreversible consequences for a claimant belong in arrangements where the system produces a recommendation with evidence and a qualified person takes the decision. That preserves the control and removes most of the effort, since assembly and analysis are usually the expensive part.
How do you tell required manual work from unnecessary manual work in claims?
Ask why each step exists and whether the reason still holds. Three categories emerge: required by regulation, required because the consequence is irreversible, and created by a system limitation. The third is recoverable and is frequently described as one of the first two by people who inherited the step.
Who should be involved in reviewing a claims process?
Case handlers and assessors, not only process owners and team leads. The long-tail cases where the effort concentrates are handled by the people closest to them, and the assembly, follow-up and reconstruction work they perform appears in no documentation that a process owner would have.
The tail is the operation
Claims organizations measure the cases that close quickly because those are the ones the systems record cleanly.
The capacity question is always about the cases that do not, and answering it requires knowing how much of a handler's week goes into assembling a picture that no system holds.
See it. Fix it. Own it.