What it means
Almost every recurring business risk can be broken into a count and a cost. Frequency is the number of loss events in a period, such as twelve customer injury claims a year, while severity is the average amount each event consumes once it is settled.
Splitting a risk this way matters because the two halves move for different reasons and respond to different fixes. Better training or safety equipment usually cuts frequency, whereas deductibles, faster claims handling or contractual liability caps usually cut severity.
A single blended loss number hides which lever is actually working. In practice you build both halves from your own claims history, normally three to five years of data adjusted for how much the business has grown.
Frequency is expressed per unit of exposure, per vehicle, per employee or per $1,000,000 of revenue, so it stays comparable as the business changes size. Severity is trended upward for inflation, because a repair that cost $6,000 three years ago costs more to settle today.
The important nuance is that the method produces an average, not a worst case. Severity in particular tends to be heavily skewed, with many small claims and a handful of very large ones, so the simple average understates what a bad year looks like.
Serious users model frequency and severity as separate probability distributions and simulate thousands of possible years to see the tail. Common variants include the burning cost approach, which just divides past losses by past exposure, and splitting attritional losses from large losses so each gets its own assumptions.
Insurers also apply the method layer by layer, because the frequency of claims above $500,000 is very different from the frequency of claims above $5,000.
In practice
Real-world examples.
Example
A bakery chain with 60 stores records 12 slip-and-fall claims a year at an average settled cost of $4,000. Expected loss is 12 x $4,000 = $48,000, which the group builds into its annual insurance budget. When a new non-slip flooring rollout is proposed, the payback is measured against that $48,000.
Example
A software subscription business processes 50,000 card transactions a month and sees a chargeback rate of 0.4%, giving 200 chargebacks at an average value of $85. Expected monthly loss is 200 x $85 = $17,000. The fraud team is set a frequency target rather than a dollar target, because it can influence how many disputes happen but not how large each order is.
Example
An electronics manufacturer ships 120,000 units a year with a 1.5% in-warranty failure rate, which is 1,800 repairs at an average cost of $65. Expected warranty cost is 1,800 x $65 = $117,000, and that figure feeds the warranty provision in the accounts. Engineering is asked to cut frequency while the service network is asked to cut severity.
Formula
Calculation
Expected Loss = Frequency x Severity
A courier business runs 250 vans. Over the past five years it has averaged 40 motor claims a year, so frequency is 40 / 250 = 0.16 claims per van per year. The average settled claim is $7,500, which is the severity.
Expected annual loss = 250 vans x 0.16 claims per van x $7,500 = $300,000. The same answer comes from 40 claims x $7,500 = $300,000. Per van that is $300,000 / 250 = $1,200 a year, which is the number the finance director budgets.
If the insurer adds a 25% loading for its own expenses, profit and uncertainty, the indicated premium is $300,000 x 1.25 = $375,000. Knowing that $300,000 of that is pure expected loss tells the buyer exactly how much of the premium is genuinely negotiable.Case study
Seen in the real world.
Halberd Logistics is an illustrative, fictional regional haulier running 90 trucks. Its broker rebuilt the motor risk from five years of claims data and found 27 claims a year at an average settled cost of $9,000, giving an expected annual loss of 27 x $9,000 = $243,000. The insurer had been quoting a flat premium with no explanation, and the split immediately changed the conversation.
Because frequency was the bigger driver, Halberd fitted telematics and introduced a monthly driver scorecard rather than buying a higher excess. Two years later claims had fallen to 18 a year and average severity had eased to $8,000, so expected loss became 18 x $8,000 = $144,000, a reduction of $99,000 against the earlier figure.
The illustrative lesson is that the arithmetic told Halberd where to spend. Had it attacked severity instead, by raising its deductible, it would have kept paying for the same number of accidents while simply moving more of each one onto its own balance sheet.
Watch out
Common mistakes.
- Treating the expected loss as a prediction of next year's actual loss, when it is a long-run average that any single year can overshoot badly.
- Using raw claim counts without dividing by exposure, so a growing business looks riskier when its loss rate per vehicle or per employee has not moved at all.
- Averaging severity across a mix of tiny attritional claims and one catastrophic claim, which produces a figure that describes neither group.
Questions
People also ask.
Why separate frequency and severity at all?
Because they respond to different management actions, so keeping them apart tells you whether to prevent losses or contain them.
How much claims history do I need?
Three to five years is a normal minimum for frequency, and longer for severity, because large losses are rare enough that a short window can miss them entirely.
Does this work for risks that have never happened to us?
Only loosely, since with no internal data you must borrow industry frequency and severity assumptions, which should be treated as a starting estimate rather than an answer.
From the founder's library
Take it further with the book.
Build your financial confidence beyond this definition. Shihan's full-length guide, Accounting Fundamentals, takes the same plain-English approach and turns it into a complete, practical playbook for non-finance managers, business owners and students - with chapter-end quiz answers and presentation slides included.
25% off with code MMHQ25, applied at checkout. Priced in USD - checkout may show the equivalent in your local currency.
View the book and save 25%Related