Average claims have become more expensive. Someone suggests fraud. It is a possible explanation, but the case also says parts costs and repair duration rose in ordinary motor claims.
I would begin with those observed changes. A good working theory follows the evidence already on the page, then asks for the comparison that could confirm or overturn it.
This fictional example shows how to move from an aggregate increase to a useful cost investigation.
Separate frequency from severity
The insurer reports:
| Measure | Supplied observation |
|---|---|
| Average cost per claim | Up 14% |
| Claims per insured exposure | Unchanged |
| Share of claims flagged for possible fraud | Barely changed |
| Standard repair claims | Parts costs and repair duration increased |
Frequency describes how often claims occur relative to exposure. Severity describes their average cost. Stable frequency does not explain a rise in cost per claim.
It also does not automatically mean the total number of claims is unchanged. If the insured book grows, claim count can grow at the same frequency. Keep exposure, count and cost per claim separate when moving to total claims expense.
Choose the stronger first hypothesis
The evidence points towards repair-related costs as the first branch to examine. That is an analysis priority, not a declaration that fraud is impossible.
A stable flag share tells you about the detection system’s output. It does not by itself establish how much fraud occurred or whether the system’s detection performance changed.
You can answer directly:
“I would investigate repair costs first because the supplied changes appear in standard claims. I would compare similar vehicles and repairs over time, separating parts, labour and costs created by longer repair duration.”
Test vehicle complexity, parts costs and repair delays against the client's claims records. The mechanism is plausible, but those records must explain this client's 14% increase.
Quick Business Judgment Drill
A pet insurer's average payout per claim rose 13% this year. The claims director suspects organised fraud and wants a dedicated investigation team.
- Claims per policy
- Unchanged year on year
- Suspicious-claim flags
- Steady at 3% of claims
- Standard vet fees
- Up 10% in unflagged claims
Which explanation should the team test first, given these facts?
Ask for comparable claims, then build a bridge
Suppose the next exhibit compares the same repair category and vehicle group across matched periods. It gives these illustrative average costs:
| Cost per claim | Earlier | Current | Change |
|---|---|---|---|
| Parts | £2,000 | £2,300 | +£300 |
| Repair labour | £1,000 | £1,100 | +£100 |
| Replacement vehicle during repair | £500 | £700 | +£200 |
| Other claim costs | £500 | £460 | −£40 |
| Total | £4,000 | £4,560 | +£560 |
The increase is £560 ÷ £4,000 = 14%. Parts explain £300 of the increase, replacement-vehicle cost £200 and labour £100, partly offset by £40 elsewhere.
You now have an explanation of the movement in this comparable group. The next question should investigate the largest actionable components, rather than return to a broad list of possible causes.
Split price from quantity within the cost line
Replacement-vehicle spending can rise because the daily rate increased or because the vehicle was needed for more days.
Suppose the daily rate stayed at £50 while average duration rose from ten to fourteen days:
10 × £50 = £500; 14 × £50 = £700.
The £200 increase comes from four additional days, so negotiating the daily rate is only one possible response. Investigate the reason for the longer repair cycle: parts availability, scheduling, assessment, authorisation or work waiting between stages.
Similarly, labour could rise because hourly rates increased, more hours were required or the mix of specialist work changed. Splitting the line identifies the intervention that fits the cause.
Quick Business Judgment Drill
A home insurer's spending on temporary accommodation for families displaced by water leaks has jumped. Procurement wants to renegotiate hotel rates.
| Year | Nights per claim | Nightly rate |
|---|---|---|
| Last year | 20 | £90 |
| This year | 26 | £95 |
How much of the increase per claim comes from extra nights, valued at last year's rate?
Check whether claim mix tells a different story
An average can also rise without any within-category cost increase. Imagine minor claims cost £2,000 and major claims £6,000 in both periods.
At a 50/50 mix, average severity is £4,000. If major claims rise to 64% of claims, the average becomes:
36% × £2,000 + 64% × £6,000 = £4,560.
That is the same 14% aggregate increase, produced entirely by mix in this separate illustration.
This is why “compare similar claims” belongs in the original data request. A category-level cost bridge and a claim-mix analysis answer different parts of the question. The weighted-average guide explains how to separate them.
Choose a next action with a measurable consequence
Return to the comparable repair group. Suppose it contains 10,000 claims a year, and a scheduling change could reduce replacement-vehicle duration by two days without compromising the repair process.
At £50 per day, the potential reduction is:
10,000 × 2 × £50 = £1 million per year.
That gives the investigation a commercial target. If the scheduling change costs £200,000 annually, the simplified net benefit is £800,000, provided those two days are actually removed from paid replacement use.
Do not count two days saved from an internal milestone if the customer still needs the replacement vehicle for the same total time. Follow the cost through to the event that stops it.
State what would change your priority
If comparable repair costs are stable and mix explains the increase, move to the mix drivers. If suspicious high-value claims account for the movement after controlling for those factors, prioritise the relevant fraud investigation.
Your view should change because the comparison changes the explanation. You do not need to defend the first hypothesis as though it were a commitment.
A strong interim answer is:
“Repair-related costs are the leading explanation in the supplied facts. I would first separate within-category cost inflation from claim mix. If longer repair duration is material, I would trace the delay to the step that extends paid replacement use.”
Try Auto Insurance Claim Severity. Give one working theory, one serious alternative and the comparison that separates them. The hypothesis-testing guide explains that sequence in another industry.
Point of view drill
Would your argument persuade an MBB interviewer?
Auto Insurance Claim Severity
Take a position on a business decision and defend it aloud. Get feedback on your reasoning, trade-offs and use of evidence.
Frequently asked questions
Framing the cost increase
What is the difference between claim frequency and claim severity?
Frequency is how often claims occur relative to insured exposure. Severity is how much each claim costs on average. Here frequency is unchanged while average cost per claim is up 14%. The problem therefore sits in what claims cost, not in how many arrive.
Why not start by investigating fraud?
The evidence on the page points elsewhere. Parts costs and repair duration rose in standard repair claims, while the share flagged for possible fraud barely changed. Starting with repair costs is an analysis priority, not a claim that fraud is impossible. It simply follows the observed changes first.
Does a stable fraud-flag share prove fraud has not increased?
No. The flag share describes what the detection system produced, not how much fraud actually occurred. If detection performance changed, a stable share could hide more or less underlying fraud. Treat it as a reason to place fraud lower in the queue, not as proof that rules it out.
Can average severity rise even if no individual repair costs more?
Yes, through a shift in claim mix. Suppose minor claims cost £2,000 and major claims £6,000. A 50/50 mix averages £4,000. If major claims rise to 64% of claims, the average becomes £4,560. That is the same 14% rise, with no change in the cost of either claim type.
Breaking down repair costs
Why compare matched repair claims rather than all claims?
Matching the same repair category and vehicle group across periods separates genuine cost changes from shifts in the kinds of claims that arrive. Without that control, a mix change could look like rising repair costs or could hide them. The comparable group lets you attribute the increase to specific cost lines.
What does a cost bridge add to the analysis?
It shows which cost lines drive the increase within comparable claims. In the illustrative exhibit, parts rise from £2,000 to £2,300 and repair labour from £1,000 to £1,100. Parts therefore contribute three times as much as labour, which tells you where deeper questions are most worth asking.
How do I tell whether replacement-vehicle costs rose because of price or duration?
Split the cost into the daily rate multiplied by the number of days. With the rate held at £50, ten days cost £500 and fourteen days cost £700. The whole £200 increase therefore comes from four extra days, not from a higher price per day.
If replacement vehicles are the issue, should we just negotiate the daily rate?
Only if the rate is what actually moved. When the increase comes from longer rentals, a cheaper rate treats the symptom while vehicles stay out for longer. Investigate why repairs now take longer, because shortening the duration addresses the real driver of the extra cost.
Turning findings into a decision
How do I put a value on reducing repair duration?
Multiply the claims affected by the days saved and the daily rate. For 10,000 comparable claims a year, cutting replacement-vehicle duration by two days at £50 a day gives a potential reduction of £1 million a year. That figure turns an open investigation into a target the business can test.
Is a scheduling fix worth it if it costs £200,000 a year?
On the simplified numbers, yes. A £1 million reduction less £200,000 of annual cost leaves roughly £800,000. However, the saving only counts if the shorter duration does not compromise the repair process. Check repair quality alongside the reduction in rental days.
What should happen if claim mix explains the increase?
If comparable repair costs are stable and a shift towards major claims accounts for the rise, move the investigation to the mix drivers. The question then becomes why more claims are severe, rather than why each repair costs more. That calls for a different set of data.
When would fraud become the priority?
Fraud becomes the priority when suspicious high-value claims account for the movement after you have controlled for repair costs and mix. At that point the evidence supports a targeted fraud investigation. The order of testing reflects the evidence available, not a judgement that fraud was never plausible.
Do I need to defend my first hypothesis if the data disagrees?
No. Your first hypothesis is a working theory, not a commitment. Say which comparison would change your view, then update when it does. A view that changes because the comparison changed the explanation is a strength of the analysis, not an inconsistency.