What it means
The core of actuarial analysis is separating frequency from severity. Frequency asks how often a claim or event occurs in a given population, severity asks how much each one costs, and multiplying the two gives the expected cost per policy or per member.
That expected cost is only the starting point. Insurers and pension schemes then load it for administration expenses, commission, capital costs and a margin for the risk that experience turns out worse than the average, which is how a raw expected cost becomes a quoted price.
Businesses outside insurance use the same discipline more often than they realise. Warranty provisions, self-insured health plans, extended service contracts and even customer refund reserves all rest on the same frequency-times-severity logic applied to internal data.
Credibility is the nuance that trips people up. A small dataset produces a wildly unstable estimate, so actuaries blend a company's own experience with broader industry data, weighting the company's own numbers more heavily as the volume of data grows.
Good actuarial analysis is also explicit about its assumptions and tests how sensitive the answer is to each one. A pricing model that only works if claim inflation stays at 3% is not really an answer; it is a bet that needs to be stated out loud.
The mix of business deserves as much attention as the headline average. A book that has quietly shifted towards larger customers or riskier trades can show stable overall claim numbers while the cost per policy underneath has moved a long way.
Splitting the analysis by segment, region or policy size is what stops that drift from going unnoticed until renewal.
In practice
Real-world examples
Example 1
A regional insurer repricing its small business property book finds claim frequency has drifted from 4% to 5%. Actuarial analysis shows the loss cost per policy rising from $240 to $300, and the renewal premium is raised accordingly rather than by a flat inflation guess.
Example 2
An electronics retailer selling three-year extended warranties uses its own repair data to estimate a 9% failure rate and a $180 average repair cost. The resulting expected cost of $16.20 per warranty tells the merchandising team the minimum price at which the product is not loss-making.
Example 3
A large employer moving to a self-insured medical plan commissions actuarial analysis of five years of claims. The study sets the monthly funding rate per employee and specifies the level at which stop-loss cover should attach. The finance team uses the same output to budget for a bad year rather than an average one.
Think of it
Actuarial analysis is like being a master weather forecaster for business finances. You cannot know for certain if it will rain tomorrow, but by studying decades of climate data, you can predict the exact chance of a storm and decide whether to invest in umbrellas or sunscreen.
Formula
Calculation
Expected Loss Cost per Policy = Claim Frequency x Average Claim Severity, and Gross Premium = Expected Loss Cost / (1 - Expense and Profit Loading). An insurer analyses a book of 5,000 small business property policies. Historical data shows 4% of policies produce a claim each year, so expected claims = 5,000 x 4% = 200 claims. The average claim costs $6,000, so expected total losses = 200 x $6,000 = $1,200,000. Expected loss cost per policy = $1,200,000 / 5,000 = $240. With expenses and profit set at 25% of premium, gross premium = $240 / (1 - 0.25) = $320 per policy. Total premium for the book = 5,000 x $320 = $1,600,000.Case study
Seen in the real world
What follows is an illustrative and fictional case. Northgate Mutual, an invented small insurer, had priced its trades package policy on gut feel and competitor comparison for a decade. Premiums were $265 per policy across a book of 5,000 policies, generating $1,325,000 of income.
An actuarial analysis of claims data showed frequency of 4% and average severity of $6,000, an expected loss cost of $1,200,000, or $240 per policy. Once the fictional insurer's genuine expense ratio of 25% was applied, the technically correct premium was $320, meaning the book had been underpriced by $55 per policy and was losing money before a single storm season.
Northgate phased the increase over two renewal cycles, accepting some customer loss in exchange for a book that covered its own claims and costs. The illustrative point is that competitor pricing tells you what others charge, not what your own risk actually costs.
Watch out
Common mistakes
- Pricing from average claim size alone while ignoring how frequently claims occur, which understates the true cost per policy.
- Building an analysis on too small a dataset and treating a volatile result as a reliable answer.
- Adding a percentage loading to the loss cost rather than dividing by one minus the loading, which produces a premium that does not actually cover expenses.
Questions
People also ask
What data does actuarial analysis need?
At minimum it needs several years of claim counts, claim amounts and the exposure base, such as the number of policies or members that generated them.
Is actuarial analysis only for insurers?
No, any business carrying uncertain future obligations such as warranties, refunds or self-insured benefits can use the same techniques.
How often should an analysis be refreshed?
Most books are repriced annually, with an interim review if claim frequency, severity or the mix of business shifts materially.