A claim is notified and nobody yet knows what it will cost. Why does the system have to put a number on it immediately, and what happens to that number?

The reserve is money, not a note

When a claim is notified the insurer becomes liable for something whose amount is unknown. Accounting does not permit waiting: the liability exists now, so an estimate of it — the case reserve — is recognised now, and it flows straight into the reported result for the period.

That single fact explains most of the machinery. The number is not an internal guess an adjuster keeps in a notebook; it is a figure that moves the company's profit, appears in regulatory returns, and will be examined. A system that treats it as an editable field on a claim record has misunderstood what it is.

The vocabulary that follows is worth being precise about, because these three are constantly conflated:

• Paid — money that has left.
• Case reserve — the estimate of what remains to be paid on this known claim.
• Incurred — paid plus reserve, the current best view of the claim's total cost.

Incurred is the number people mean when they say what a claim "cost", and it changes every time either component does.

The first number is a placeholder, deliberately

At first notification almost nothing is known. Practice is to set an initial reserve from an average — a formula reserve based on claim type, or a standard opening figure by peril — and to replace it as soon as a human has looked.

That is not sloppiness, it is the least-wrong option. A zero reserve understates the liability and produces a step change later; a speculative large number distorts the period it lands in. An average by claim type is defensible, is consistent across the book, and is expected to move.

What matters for the design is that the transition from formula to adjuster-assessed is itself an event worth recording, because a reserve that moved because someone finally read the file is a different thing from one that moved because the claim got worse.

Every movement is an event

The reserve is not a value, it is a series. The current figure is derived from the movements, and the movements are what get reported.

Claim CL-8841, motor, third-party injury

date        movement          case reserve   paid      incurred   reason
2026-02-14  initial formula        3,500.00      0.00    3,500.00  FNOL, average by type
2026-03-02  adjuster estimate     12,000.00      0.00   12,000.00  medical report received
2026-05-19  payment                8,000.00   4,000.00   12,000.00  interim payment
2026-09-30  strengthening         26,000.00   4,000.00   30,000.00  liability admitted, PSLA revised
2027-01-08  settlement                 0.00  29,500.00   29,500.00  closed, settled below reserve

Read the fourth row: the reserve rose by 18,000 in one movement, which is a number that appears in that period's result and will be asked about. Read the third: a payment of 4,000 reduced the reserve by 4,000 and left incurred unchanged, because paying out of a reserve is not a new cost — it is the liability converting into cash. Candidates who model a payment as reducing only the reserve, or as increasing only paid, get incurred wrong in one direction or the other.

The reason to store movements rather than a current value is that actuaries need to see cost by accident period and development period — how a cohort of claims from one period developed over subsequent periods — and that is reconstructible from a movement history and impossible from a mutable field.

Indemnity and expense are separate

A claim costs two different kinds of money. Indemnity is what is paid to settle the loss itself. Expense is what it cost to handle it: loss adjusters, legal fees, engineers, surveyors.

They are reserved and reported separately because they behave differently and are managed by different levers. A long liability claim can run up substantial legal expense while indemnity stays flat; reinsurance treaties often treat them differently; and expense ratios are a management metric in their own right. Collapsing both into one number destroys information nobody can recover later.

IBNR covers what the case reserves cannot

Case reserves only exist for claims the insurer knows about. Claims have occurred that have not yet been reported — the accident happened, the insured has not called — and on some lines the delay is years.

IBNR is a bulk reserve for that gap, set by actuaries across a whole cohort rather than per claim. It is worth understanding as a different object with a different owner: an adjuster sets a case reserve on a file, an actuary sets IBNR on a book, and an engineer's job is to feed the second with clean movement data rather than to compute it.

The related consequence is that "total reserves" is case plus IBNR, and a report that sums only case reserves understates the liability on any line with a reporting delay.

Closed is not final

A claim can reopen — an injury worsens, a new party joins a suit, a recovery arrives — and reopening restores a reserve on a file that had been closed at zero.

Two design consequences follow. Closed claims cannot be archived into somewhere expensive to write to, because reopening is a normal event rather than an exception. And the reserve movement history has to survive closure, since the reopened file's development is part of the same claim's story and an actuary looking at development will want both.

The blunt version of the same point: on a long-tail line, an insurer does not know what a year cost for a decade, and every number before then is an estimate with a history attached.

The reserve is recognised money set against something nobody has measured yet. Model the movements, not the value, because the movements are what the business, the actuary and the regulator all actually read.