Claim severity
What is claim severity: claim severity explained, and how it is calculated
Claim severity is the average cost of a claim, calculated by dividing the total amount paid over a period by the number of claims in that same period. It is one half of the pair that drives loss cost, and it answers a single question: when a claim happens, how expensive is it?
For insurers
See the damage before you send anyone
Venta Capture, a product of VentaVid, sends the policyholder a link. They film the damage on their own phone, guided step by step, and the evidence lands with the claim.
You will also see it written as average claim severity, average claim cost, or loss severity. Insurers usually measure it per coverage rather than across a whole book, because a comprehensive claim and a bodily injury claim are not comparable numbers.
How is claim severity calculated?
The formula is simple. Getting the inputs right is where the work sits.
Claim severity = total losses incurred, divided by the number of claims.
Three definitional choices change the answer, so agree them before anyone compares two figures:
- Paid or incurred? Paid severity counts money that has left the building. Incurred severity adds outstanding reserves, so it reflects open files as well as closed ones.
- Does it include loss adjustment expense? Some measures carry the cost of handling the claim, some do not. Say which.
- Which claims count? Closed-with-no-payment claims sit in the denominator on some definitions and not on others, which alone can move severity by a wide margin.
Severity is also a lagging measure with a long tail. On injury claims especially, the severity of an accident year keeps moving for years as reserves develop.
How claim severity differs from claim frequency
Frequency is how often claims occur, usually expressed as claims per 100 units exposed to risk per year, for instance per 100 insured vehicles. Severity is how much each one costs. Multiply them and you get pure loss cost per unit of exposure, which is the number pricing actually cares about.
They move independently, and that is the point of tracking both. US private passenger auto data compiled by ISO, a Verisk business, shows property damage liability frequency falling from 3.50 claims per 100 car years in 2012 to 2.28 in 2021, while severity over the same period rose from $3,073 to $5,314 per claim (Insurance Information Institute). Fewer claims, each one substantially more expensive.
Reading either number on its own leads somewhere wrong. A falling claim count looks like good news right up to the point where the average cost has quietly doubled.
Claim severity: a worked example
A motor book pays out $12.4 million across 2,300 claims in a year. Severity is $5,391 per claim. The following year the same book pays $12.9 million across 2,050 claims, so severity is $6,293. Total spend rose 4 percent, claim count fell 11 percent, and average claim cost rose 17 percent. Nothing about the frequency number would have told you that.
What pushes severity up
- Repair complexity. Sensors, cameras, calibration steps and expensive panels raise the cost of damage that used to be cheap. CCC Intelligent Solutions' Crash Course 2026 report puts the average auto repair cost at between $4,500 and $5,000 in 2025, against almost $2,500 in 2010 (reported by Claims Journal).
- Parts and labour inflation, plus supply delays that extend hire and accommodation costs.
- Total loss thresholds. When repair costs rise against vehicle values, more claims tip into total loss, which resets severity upwards.
- Injury and legal costs, the largest and slowest-developing driver on liability books.
- Claims handling itself. Delays, rework and inconsistent scoping add cost to files that were not inherently expensive.
How claims operations use severity
Pricing and reserving teams use severity as an input. Operations use it as a segmentation tool, and that is the more practical application day to day.
- Setting triage thresholds. Severity distribution across your own book, not a market average, should decide where the desk cut-off sits. See claims triage.
- Sizing reserves early. Severity by segment gives handlers a defensible opening reserve rather than a guess.
- Allocating capacity. High-severity segments justify experienced handlers and attendance. Low-severity, high-frequency segments justify automation.
- Spotting leakage. A severity figure drifting up while the underlying mix is unchanged points at handling, not at the market.
A caution about averages
Severity is a mean, and claims distributions are heavily skewed. A handful of large losses can lift an average that describes almost none of the actual files. Look at the median and the tail alongside it, and segment before you compare anything. A single blended severity figure across a mixed book is close to useless for operational decisions.
One more thing worth watching next to severity: how long files stay open. Duration and cost move together, which is why claims cycle time belongs on the same dashboard, and why the practical routes to shortening it are covered in our guide to reducing claims cycle time.
