Training ROI is the one number a finance director will ask for and the one most training teams cannot produce with a straight face. The arithmetic is simple. The hard part is deciding what goes into each side of the calculation, and being honest about how much of the result the training actually caused.
This guide covers the formula, what belongs in the cost and benefit columns, a worked example with the numbers shown, a worksheet you can fill in for your own program, and the cases where ROI is the wrong measure altogether.
Training ROI is the net benefit of a program divided by its cost, expressed as a percentage:
Training ROI (%) = (Program benefits − Program costs) ÷ Program costs × 100
A result of 0% means the program paid for itself and nothing more. A result of 50% means every dollar spent came back with fifty cents on top. A negative result means the program cost more than it returned, at least over the period you measured.
It is worth reporting the benefit-cost ratio alongside it:
Benefit-cost ratio = Program benefits ÷ Program costs
A ratio of 1.3 says the same thing as an ROI of 30%, but some audiences find a ratio easier to read, and it avoids the confusion of negative percentages when a program falls short.
Most training ROI figures are inflated on the cost side, not the benefit side, because the costs that get counted are the ones on an invoice. A fully loaded cost includes:
Leaving out learner time is the most common way a mediocre program ends up with an impressive ROI. The worked example below shows how large that effect can be.
Benefits are harder, because they have to be observed, isolated and converted to money. Take each in turn.
Start with an outcome someone outside the training team already tracks: time to productivity, error or incident rates, first-year turnover, sales per representative, audit findings, customer complaints. If nobody tracks the outcome today, you will struggle to show that it moved.
Outcomes rarely change for one reason. A new hiring manager, a system upgrade or a seasonal swing can move the same number. The strongest way to isolate training's share is a comparison group that did not take the program. Where that is not possible, trend-line analysis against the period before the program is the next best option. The weakest acceptable method is asking managers or participants to estimate how much of the change was due to training, then discounting that estimate by how confident they say they are.
Whatever method you use, state it next to the result. An ROI figure without its isolation method is the first thing a skeptical executive will challenge.
Some outcomes carry a known cost: the replacement cost of a leaver, the average cost of a safety incident, the margin on a sale. Use the figure your finance team already uses, rather than a number from an industry report, so the calculation is built on data they recognize.
Benefits you cannot credibly convert, such as morale or confidence, are still worth reporting. Report them alongside the ROI figure, not inside it.
The figures below are illustrative. They are not drawn from a customer program.
A regional healthcare network runs an onboarding program for 150 new clinical support staff a year. Before the program, new hires took about ten weeks to reach full productivity and 30% left within their first year. After one year of the program, time to productivity has fallen to eight weeks and first-year turnover to 25%. A new scheduling system was introduced in the same year, so the network attributes 70% of the improvement to training, based on a comparison with a site that had not yet adopted the program.
| Cost | Calculation | Amount |
|---|---|---|
| Content development (year-one share) | Amortized build cost | $30,000 |
| Platform and hosting share | Allocated licence cost | $12,000 |
| Preceptor time | 150 hires × 6 hours × $45 | $40,500 |
| Learner time | 150 hires × 16 hours × $28 | $67,200 |
| Administration | 150 hours × $40 | $6,000 |
| Total cost | $155,700 |
| Benefit | Calculation | Amount |
|---|---|---|
| Faster time to productivity | 2 weeks saved × 50% productivity gap × $1,120 weekly wage × 150 hires | $168,000 |
| Lower first-year turnover | 7.5 fewer leavers × $12,000 replacement cost | $90,000 |
| Gross benefit | $258,000 | |
| Share attributed to training | $258,000 × 70% | $180,600 |
Now remove learner time from the cost column, as many calculations do. Total cost falls to $88,500 and ROI jumps to 104%. The program has not changed. Only the accounting has. A 16% return that survives scrutiny is worth more in a budget meeting than a 104% return that collapses at the first question.
Copy this table into a spreadsheet and fill it in for your own program. Keep the notes column: it is where you record the source of each figure, which is what makes the result defensible.
| Line | How to calculate it | Source or note |
|---|---|---|
| A. Development cost | Build cost ÷ years the content will be used | |
| B. Platform and content licences | Share of annual cost used by this program | |
| C. Delivery | Facilitator hours × loaded rate, plus travel and venue | |
| D. Learner time | Learners × hours in training × loaded hourly wage | |
| E. Administration and evaluation | Hours × loaded rate | |
| F. Total cost | A + B + C + D + E | |
| G. Outcome change | After minus before, in the outcome's own units | |
| H. Value per unit | The figure your finance team uses | |
| I. Gross benefit | G × H, summed across outcomes | |
| J. Share attributed to training | Percentage, with the isolation method named | |
| K. Net benefit | I × J | |
| ROI | (K − F) ÷ F × 100 | |
| Benefit-cost ratio | K ÷ F |
ROI sits at the top of a ladder, and it cannot be built without the rungs underneath it. The Kirkpatrick model describes four levels of evaluation: reaction (did learners find it useful), learning (did they gain the knowledge or skill), behavior (are they doing things differently on the job) and results (did a business outcome change). The Phillips model adds ROI as a fifth level, reached by converting level-four results to money and comparing them with cost.
The practical consequence is that an ROI study is only as strong as the evidence at level three. If you cannot show that behavior changed, any business result you attribute to training is a guess. Measuring training effectiveness at the lower levels is not a lesser alternative to ROI. It is the evidence ROI depends on.
That evidence usually lives in more than one system: completions and assessment scores in the LMS, role and tenure data in the HRIS, outcomes in operational or finance systems. Joining them by hand is where most ROI studies stall. A learning analytics platform exists largely to do that join, which is worth knowing before you decide whether to build the analysis in spreadsheets or buy something to do it.
Some programs should not be judged on ROI at all.
Mandatory compliance training. The return on a required program is the avoidance of a penalty, a finding or an incident, and the honest counterfactual is not "no training" but "non-compliance." Report currency instead: the share of the required population holding a valid credential today, and how that has changed. That is the figure a regulator or auditor will ask for.
Programs too small to measure. If the cost of a credible ROI study approaches the cost of the program, measure at levels one to three and stop there.
Outcomes with a long lag. Leadership development and culture programs may take years to show up in business results. Agree in advance which leading indicators will stand in for ROI in the meantime, rather than producing a first-year figure that understates the program.
Whichever measure you use, the way you present it matters as much as the number. Our guide to presenting training impact to senior leadership covers how to translate program data into the terms executives use.
There is no reliable universal benchmark. A more useful test is whether the program clears the same hurdle your organization applies to other investments. If finance expects capital projects to return 15%, a training program that credibly returns 16% is holding its own.
Yes. Time spent in training is time not spent on other work, and leaving it out routinely overstates ROI by a wide margin, as the worked example above shows.
Long enough for the outcome to move and for the change to be distinguishable from normal variation. For onboarding or skills programs that is often six to twelve months. Measuring too early understates the return; measuring too late makes it harder to isolate training's effect.
ROI converts results to money and compares them with cost. Return on expectations asks whether the program delivered the outcomes its sponsors agreed to at the start, whether or not those outcomes can be monetized. Many organizations use expectations for programs where a credible dollar figure is out of reach.
Not on its own. An LMS holds the cost-side activity data and the learning evidence, but the business outcomes that make up the benefit column live in other systems. Reporting tools such as Zoola Analytics can bring LMS data together with those sources, but deciding what counts as a benefit and how much of it training caused remains a judgment your team has to make and defend.