Expected losses, the primary and excess split, the ERA, weight and ballast, worked claim by claim on one employer, then converted into dollars of premium.
The experience mod is the factor that multiplies a workers' compensation premium by the employer's own loss history. A mod of 1.00 is average; 1.20 means a 20 percent surcharge on the rated premium, and a mod below 1.00 earns a credit. The rating bureau computes it, but the formula is public and every input sits on the worksheet the bureau sends you, so the mod can be rebuilt, checked and projected in a spreadsheet. This guide walks the NCCI formula end to end on Brannock Supply Co., a fictional regional distributor with 520 employees, five classes and 84 claims in its experience period. Its January 1, 2027 mod comes out at 1.1980, rounded 1.20, and that one factor costs it $154,125 a year.
Worked in full in The Workers' Compensation Handbook by Julian R. Sterling, with every figure reproduced in live Excel workbooks.Get the DRM-free PDF, $19.99 →
Under the NCCI experience rating plan, used in most states, the mod is:
mod = (Ap + W x Ae + (1 - W) x Ee + B) / (E + B)
E is expected losses, Ep the expected primary losses and Ee the expected excess losses (E minus Ep). Ap and Ae are the actual primary and actual excess losses. W is the weight and B the ballast, both read from the state's table according to the size of E. The result is rounded to two decimals, and the premium uses the rounded value.
The experience period is normally the three policy years ending one year before the rating effective date. For Brannock's January 1, 2027 rating that means policy years 2023, 2024 and 2025; 2026 is still too immature to be valued and stays out. California and New York run their own rating plans, so the steps below describe the NCCI plan in principle, as read in November 2026. States differ in their values, and a few in their rules.
Expected losses are what an average employer of the same size and class mix would have lost over the experience period. For each class and year, payroll is divided by 100 and multiplied by the expected loss rate (ELR) the bureau publishes for that class. Each expected loss is then split by the class D-ratio, the share of losses an average employer would report below the split point: E x D-ratio gives the expected primary losses.
One line shows the mechanics. In 2025, Brannock's trucking class 7219 carried payroll of $5,362,000. At an ELR of 2.45, E = $5,362,000 / 100 x 2.45 = $131,369. With a D-ratio of 0.30, expected primary losses are $39,411. Repeat for every class and every year and add up.
| Class | 3-year payroll | E | Ep |
|---|---|---|---|
| 8810 Clerical | $16,544,000 | $11,580 | $4,401 |
| 8742 Outside sales | $8,272,000 | $11,580 | $4,169 |
| 8018 Wholesale store | $30,420,000 | $358,956 | $122,045 |
| 7219 Trucking | $15,476,000 | $379,162 | $113,749 |
| 3632 Machine shop | $11,475,000 | $174,420 | $57,559 |
Total expected losses are $935,698, of which $301,923 are expected primary and $633,775 expected excess. Two classes, the branches and the trucks, make up most of the expected side, which is where the actual claims will be judged.
Actual losses are taken claim by claim from the unit statistical reports, and three adjustments are applied before anything is added.
| Claim | Incurred | Primary | Excess |
|---|---|---|---|
| 23-L03, machine shop | $9,800 | $9,800 | $0 |
| 23-L08, trucking | $21,700 | $20,000 | $1,700 |
| 25-L01, branch (open) | $85,000 | $20,000 | $65,000 |
| 23-L04, rollover (open) | $312,000 | $20,000 | $240,000 |
The table already tells most of the story. The $21,700 claim and the $85,000 claim put exactly the same primary amount on the worksheet. Across all 84 claims, actual primary losses Ap are $508,626 and actual excess losses Ae are $599,900, so the $1,284,020 of incurred losses becomes $1,108,526 of rated losses after ERA and limitation.
The split point itself is no longer uniform. With ratings effective from November 1, 2023, NCCI replaced the single countrywide split point of $18,500 with state-specific values, reported to range from $9,500 in Oregon to $38,000 in Louisiana, and rebased the limitations on each state's largest claims. A mod trend line that crosses that date needs a footnote.
The weight decides how much of the employer's own excess record to believe; the rest of the excess side is filled with expected excess. The ballast is a fixed amount added above and below the line, which pulls small employers toward 1.00. Both grow with expected losses and come from the state's table. Brannock's E of $935,698 falls in the case state's row from $500,000: W 0.17 and B $79,000. The next row, from $1,000,000, would give W 0.24 and B $104,000, so an employer whose payroll grows will see its excess losses count for more.
Put the pieces in the formula:
The free Experience Mod Calculator reproduces exactly these cells, prefilled with Brannock, so you can watch each input move the result before replacing them with your own worksheet.
A mod is more useful decomposed than quoted. Compare each actual figure with its expected counterpart: Brannock's actual primary losses are 1.68 times expected primary ($508,626 against $301,923), while actual excess is 0.95 times expected excess ($599,900 against $633,775). The debit comes from frequency, the steady run of 28 claims and eight lost-time claims a year, most of which reach the split point, and not from the rollover.
Three sensitivities confirm it:
That last point is the practical rule for operations: preventing a claim moves the mod far more than shortening it.
The mod multiplies the premium after the manual rates and increased limits, before the schedule credit and the premium discount. On Brannock's 2027 policy the guaranteed cost premium is $936,603 at the 1.20 mod and would be $782,478 at a 1.00 mod: the debit costs $154,125 a year. Each 0.01 of mod moves standard premium by $8,476, and the extra $60,000 claim above raises the annual premium by $15,412, about $46,237 over the three ratings in which it stays.
Multiplying mod points by standard premium overstates the effect, because the premium discount and the state assessment act on the change before it reaches the invoice. The full algorithm, and how a loss-sensitive plan carries the mod inside its own formula, is covered in the guide on how a retrospective rating plan is priced.
Once you can rebuild the mod, review the bureau's worksheet line by line. Payroll by class should match the audited payroll of each policy year, which is why what a premium audit changes matters to the mod as much as to the premium. Medical-only claims must carry the ERA flag; a claim miscoded as lost-time enters at full value. Open claims should carry reserves that reflect the file: had claim 25-L01, open at $85,000, been reviewed and reserved at $40,000 before the valuation, the mod would have been 1.1905, rounded 1.19, and the annual premium $928,896 instead of $936,603, a saving of $7,706. The reduction sits entirely in excess losses, weighted 0.17, which is why reserve reviews matter less to the mod than claim counts.
Finally, project the next rating. For January 1, 2028, policy year 2023 drops out with the rollover and 2026 enters; on the book's assumptions the mod falls to 1.0628, rounded 1.06. That improvement is a roll-off, not a performance gain, and a budget should say so. The complete worksheet review, the three-year mod budget and the claims side are worked in The Workers' Compensation Handbook.
Below 1.00 earns a credit and above 1.00 a debit, so a lower mod is better. What matters more is why the mod sits where it does. Brannock's 1.20 comes from frequency: actual primary losses are 1.68 times expected while excess losses are 0.95 times expected.
Primary losses, the first dollars of every claim up to the split point, count at full value, while excess losses count only at the weight. With a $20,000 split point and a 0.17 weight, a $20,000 claim adds $20,000 to the numerator and a $60,000 claim adds $26,800.
Normally three ratings, because the experience period covers three policy years ending one year before the rating date. One extra $60,000 claim costs Brannock about $15,412 of premium a year, roughly $46,237 over its life in the mod.
Yes, if you have the bureau's worksheet: payroll by class, ELRs, D-ratios, your state's split point, limitation, weight, ballast and the claim list. The free Experience Mod Calculator does the arithmetic. The bureau's worksheet remains the document of record.
This guide is one arc of The Workers' Compensation Handbook. The book follows one employer through class codes and payroll, the premium algorithm, the experience mod claim by claim, the premium audit, claims and return to work, the loss development triangle and the choice between guaranteed cost, a retrospective rating plan, a large deductible and self-insurance.
Paperback PDF edition, $19.99 Book page and companion workbooks
Free download: Experience Mod Calculator (Excel).
Illustrative figures computed by the book's model; the company, its claims and the case state's rating values are fictional. Rules described in principle as read in November 2026; states differ. Education only, not insurance or legal advice.