If you run a construction, trades, or manufacturing business, you probably already know more about workers compensation than the average CEO. You know your premium is driven by two things: which class codes your employees fall into, and how much payroll runs through each class. You audit your codes. You push back when the carrier misclassifies a foreman as a roofer. You watch overtime.

That’s smart. But it’s only two-thirds of the picture.

The third lever — and in most cases the biggest one — is your experience modification factor. Your ex-mod. And almost nothing about your ex-mod is determined by your insurance broker, your carrier, or your accountant.

It’s determined by how you run your shop.

Which means it’s an HR problem. And if you want to lower workers compensation premium, the path runs straight through HR infrastructure you probably don’t have yet.

What the ex-mod actually is

Here’s the short version. Every year, a rating bureau (NCCI in most states, a state bureau in California, Pennsylvania, and a handful of others) compares your actual claims losses to the expected losses for a business your size in your industry.

The ratio becomes a multiplier on your premium.

A 1.0 means you’re average. You pay manual premium.
A 0.75 means your losses run 25% under expectation. You pay 25% less.
A 1.35 means your losses run 35% over. You pay 35% more.

Same payroll. Same class codes. Wildly different premium.

Two more things worth knowing, because they change how you should think about claims:

Frequency hurts more than severity. The formula deliberately puts more weight on the number of claims than on the dollar size of any one claim. Ten small claims will move your mod more than one big one. That sounds backwards, but it’s intentional — the bureau is trying to predict your future, and frequency is a better predictor than the random luck of one bad accident.

The look-back is three years, dropping the most recent. Your 2026 mod uses claim data from roughly 2022, 2023, and 2024. Which means the work you do this year shows up two years from now. Start late and you’re paying for the past for a long time.

Why this is an HR problem, not an insurance problem

When most owners decide to “do something” about workers comp, they call their broker. Their broker shops the market. Maybe they save a few points.

What the broker can’t do is change the number sitting on your experience rating worksheet.

That number is built from:

How often your employees get hurt
How quickly hurt employees get back to work (or don’t)
Whether injuries get reported promptly and accurately
Whether your supervisors know what to do in the first 24 hours after an incident
Whether you have documentation that defends against fraudulent or inflated claims
Whether your training is real or theoretical

Every single one of those is HR infrastructure. Not insurance. Not safety glasses. Not posters in the breakroom.

This is the part most owners miss. They treat workers comp as a line item on the P&L that gets negotiated once a year. It’s actually a trailing indicator of how well your HR function is running. If your HR function is your office manager doing it between phone calls, your ex-mod is going to reflect that for years.

The four pieces of HR infrastructure that move your ex-mod

Here’s the practical half. If you build these four things — honestly, not just on paper — you will see your ex-mod come down. Not next quarter. But on a three-year horizon, materially.

1. A real incident reporting process

Most small businesses don’t have one. They have a guy named Mike who handles it.

A real process means:

Every incident, no matter how small, gets documented the same day. Not just the ones that look like they’ll turn into claims.
There’s one form (or one digital tool) every supervisor knows how to use.
Reports include witness names, time, location, what happened, what the employee said, and what was done.
The owner or HR lead sees every report within 24 hours.

Why this matters for your mod: incomplete documentation is what turns a $1,500 medical-only claim into a $35,000 disputed indemnity claim with a lawyer attached. And remember — frequency matters more than size, but undocumented incidents become both frequent and expensive.

2. A written return-to-work (RTW) program

This is the single highest-leverage thing on this list. It’s also the one almost nobody actually has.

Here’s why it matters in plain numbers. In the ex-mod calculation, medical-only claims get reduced by 70% before they hit your mod. Indemnity claims — the ones where the employee is out of work and collecting lost wages — go in at nearly full weight.

So a back strain where the employee misses three weeks of work is a different animal than the same back strain where the employee is back on light duty in four days running the parts counter or doing paperwork.

A real RTW program means:

Modified-duty job descriptions written before you need them — for every department.
A standing relationship with an occupational medicine clinic that understands you’ll take employees back on restrictions.
A policy that says “if you can come to work, you come to work, even if it’s not your normal job.”
Supervisors trained to actually use the modified-duty roles instead of telling the employee to stay home and rest.

You will not get this from your insurance carrier’s “loss control” packet. You’ll get a binder. The binder is not the program.

3. Documented training that you can prove happened

OSHA training, equipment training, safety meetings — most shops do these. Most shops cannot prove they did them.

For your ex-mod, training mostly matters in two ways: it reduces claims frequency (the biggest mod driver), and it gives you defensible documentation when a claim gets contested.

What “documented” actually means:

Sign-in sheets for every training, with date, topic, instructor, and signatures.
Records kept by employee, not just by date — so when Mike gets hurt three years from now, you can pull his entire training history in five minutes.
New-hire onboarding that includes safety training before the employee touches a tool, with a signed acknowledgment.
An annual refresher schedule that runs on a calendar, not when someone remembers.

This is exactly the kind of thing that gets dropped when HR is a part-time job for someone whose actual job is something else.

4. A hiring and supervisor process that screens for the things that drive claims

Two things drive most workers comp claims in physical-labor businesses: people who shouldn’t be in those jobs, and supervisors who run their crews badly.

You can’t screen for either of those with a job application and a handshake.

The infrastructure piece looks like:

Written job descriptions that include actual physical demands — lifting weights, repetitive motions, standing hours. Used in hiring conversations, not just stored in a file.
A pre-employment process that’s legally compliant but actually probative. Post-offer physicals where appropriate. Reference checks that ask real questions.
Supervisor training on what they’re allowed to say and do when an employee gets hurt. (Hint: “walk it off” is the wrong answer, and it shows up in depositions.)
A clear chain of command for the first 24 hours of any incident.

A surprising amount of claim inflation happens in those first 24 hours. A supervisor who panics, or one who blows it off, can each turn a manageable incident into a six-figure claim.

Where this goes wrong

The most common failure mode is buying a binder.

Owner reads an article like this one, decides he needs an RTW program, calls his broker, and the broker sends over a template. The template gets printed, put in a binder, and lives on a shelf. Nothing changes operationally. The mod doesn’t move.

The infrastructure is not the document. The infrastructure is the document plus the operating habits that make the document real — supervisors who know it exists, modified-duty roles that actually get used, documentation that actually gets filed.

The second failure mode is starting and stopping. Owners get fired up, do it for a quarter, then drift. The ex-mod calculation runs on three years of data. You don’t get to drift.

The third is treating it as the office manager’s problem. Your office manager has a full-time job already. HR is not the spare time between phone calls.

What this is worth

Run the math on your own business. Take your annual workers comp premium. Multiply it by 0.20. That’s roughly what the difference between a 1.10 mod and a 0.90 mod is worth to you every year.

For a lot of construction and manufacturing shops, that’s a five- or six-figure number. Year after year.

It also compounds. A lower mod makes you more competitive on bid jobs that require a mod under 1.0 to even prequalify. In some industries that’s the difference between getting on the bid list and not.

The work to get there isn’t glamorous. It’s incident forms and modified-duty job descriptions and training sign-in sheets. It’s the kind of work that gets done by an HR function that actually exists, run by someone who actually knows what they’re doing.

If your mod is over 1.0 and you’ve never thought about it as an HR problem, let’s talk. The longer you wait, the longer you pay for it.