Every CFO eventually asks the same question about the training line item: what did we get for it? The answer that usually comes back is attendance numbers and a satisfaction score. Neither number survives a budget review, and neither should.
There is a better answer, and it has existed for decades. Jack Phillips built a five-level evaluation model that most learning and development departments already know by name: reaction, learning, application, business results, and financial ROI. The trouble is that almost nobody finishes the climb. Training Industry’s own review of the Phillips ROI Methodology notes that only 5 to 10% of corporate training programs are ever evaluated at that fifth level, the level that turns training into a dollar figure a CFO can compare against anything else competing for the same budget. In terms of workers’ compensation, where the training line item competes against the technology spend, headcount, and reserves, that gap is expensive. It means most claims organizations are running a training program and hoping it works, without ever finding out the utility or efficacy.
Timing makes the gap harder to ignore. WCRI’s CompScope Benchmarks, 2026 Edition, shows total claim costs rising at 6% a year across the median of its 18 study states from 2022 through 2025. That is not a one-year spike. It is a multi-year trend across medical payments, indemnity, and benefit delivery expense. A training program that cannot demonstrate its effect on that trend is not neutral. It is a cost sitting next to a rising cost, unaccounted for either way.
Start the model with a realistic number for what the training itself costs. ATD’s 2026 State of the Industry report put direct training expenditure at $846 per employee in 2025, down sharply from $1,254 the year before, even as formal learning hours rose from 13.7 to 16.7 per employee. A one-year swing that large in the dollar figure is more likely a reporting change than a true shift in spending, so treat the number as a starting anchor rather than gospel. The more reliable signal is the second half of that sentence. Organizations are trying to teach more on a tighter budget, which is exactly the discipline a 12-month ROI model is built to support.
The model itself follows Phillips’ levels in calendar order, because claims data develops slowly and the levels happen to unfold in roughly the same order the data becomes available.
Months 1 and 2 establish the baseline. Before a single adjuster sits through a session, pull the claims scorecard by cohort: average claim duration, reopen rate, litigation rate, reserve accuracy, and medical cost trend. This step gets skipped more often than any other, and skipping it is what turns a training program into an anecdote instead of a measurement.
Months 3 and 4 are delivery and comprehension, Phillips’ levels one and two. Test what adjusters actually retained, not how they felt about the room or the instructor. A satisfaction score of 4.8 out of 5 does not appear on anyone’s income statement.
Months 5 and 6 move to level three, behavior on the file. Audit a sample of claims handled by trained adjusters. Is mechanism of injury documented before advanced imaging is ordered? Is peer review referred earlier when a diagnosis outruns the objective findings? This is where most programs stop measuring, and it is also where most of the eventual financial value is still invisible on a P&L.
Months 7 through 9 are level four, business results starting to surface. Rerun the same scorecard from months 1 and 2. Reopen rate and reserve accuracy tend to move first because they reflect decisions made close to the point of training. Litigation rate and claim duration lag behind, because those outcomes depend on how a claim was handled months earlier.
Months 10 through 12 are level five, the financial calculation itself. Compare the trained cohort’s scorecard to its own baseline, and if the organization is large enough to support it, to an untrained comparison cohort managing a similar mix of claims. Multiply the measured movement by claim volume and by a defensible average cost per claim.
A worked example keeps the model honest. Take a mid-size TPA with 50 adjusters handling 6,000 claims a year. Training investment at the ATD-anchored figure runs 50 times $846, or $42,300 for the year. The National Safety Council’s Injury Facts data, sourced from NCCI’s Workers Compensation Statistical Plan, puts the average cost across all claims at $47,316. Multiply that by 6,000 claims and the book represents roughly $284 million in claims exposure for the year. Against that base, the entire training line-item breaks even at a movement of about 0.015% in average claim cost across the book. That is not a bet on a large effect. It is a very small ask of the data; smaller than the quarter-to-quarter noise most claims organizations already see in their own numbers.
That does not mean every improvement in the scorecard belongs to the training program. Reserve philosophy changes, panel changes, litigation posture, and the same macro severity trend from WCRI all move at the same time. A rigorous 12-month model accounts for that by comparing trained and untrained cohorts side by side, or by staggering the rollout so some offices train in month three and others in month nine, giving the model an internal control group instead of a single before-and-after snapshot.
This does not require perfect attribution to be useful. It requires a documented baseline, a known training cost, a known claim volume, and a scorecard rerun at regular intervals rather than reviewed once. Built this way, the 12-month model answers the CFO’s question in the CFO’s own language: dollars measured against a number the CFO already follows.
For claims professionals who want to build fluency in the clinical side of that scorecard, particularly how mechanism of injury and objective findings should drive documentation and cost, WorkCompCollege.com offers certification coursework built for exactly that gap.
Sources
- Phillips ROI Methodology, Training Industry glossary, trainingindustry.com/glossary/phillips-roi-methodology.
- Workers Compensation Research Institute, CompScope Benchmarks, 2026 Edition, wcrinet.org.
- Association for Talent Development, 2026 State of the Industry Report, td.org.
- National Safety Council, Injury Facts, Workers’ Compensation Costs, sourced from NCCI Workers Compensation Statistical Plan database, injuryfacts.nsc.org.
