The slide claims a strong return, and the finance director reads it twice before asking the one question the deck skipped: what would have happened without the training? Measuring training outcomes in money only works once you separate what the program caused from everything else that changed, and then load the full cost against it.
A number built that way survives the question, while one assembled the night before usually does not. You will see why the quick version fails, three ways to isolate the training’s effect, how to turn it into money and what to count as cost. The worked example, start to finish, puts every step on one table.
TL;DR
Training ROI is net benefit divided by fully loaded cost, multiplied by 100, and the arithmetic is the easy part. The hard part is the benefit, because a raw before-and-after gain credits the training with everything else that moved at the same time. First isolate the training’s share with a comparison group, a trend line or a participant estimate discounted by the person’s stated confidence. Then convert that share to money with a value finance already uses, an expert’s estimate or the participant’s own discounted figure. Load every real cost, including wages for time off the job, administration and the evaluation itself. Count one year of benefit for a short program, round down when unsure and print the isolation assumption next to the final figure.
What Does the ROI Number Compare?
The ROI number compares what a program returned, after its costs, with what it cost. The formula is net benefit divided by cost, multiplied by 100, and it sits on top of a longer chain of evidence.
Jack and Patricia Phillips, whose ROI Methodology is widely used in training evaluation, add ROI as a fifth level on top of Kirkpatrick’s four. Their 2005 paper on coaching gives a simple case: a coaching project with ten people produced $288,000 in benefits against $74,000 in costs, a return of 289%. The paper and its figures were re-read on October 6, 2026.
That figure needed one judgment before any arithmetic: how much of the $288,000 the coaching caused rather than the quarter it ran in. Every honest ROI figure rests on a judgment like that, which is why the steps below matter more than the formula.
The first four levels, from reaction through business impact, have to be evidenced before the fifth means anything. Keeping that evidence is the job of measurement tools for coaches and L&D teams, and a percentage with no record behind it has nothing to stand on.
The Naive Number, and Why It Fails a Second Question
The quick version of training ROI credits the whole gain to the training, and it fails the moment someone asks what else changed. Here is how it usually looks.
Suppose a retailer trains 25 sales staff in selling technique. Before training, each sold two garments a day at $20 each, and afterward each sells four, a gain of $40 a day per person. Across 25 people and 100 trading days that is $100,000, against a training cost of $25,000, so the report shows a net benefit of $75,000 and an ROI of 300%.
Every number in that sum is arithmetically correct, and the conclusion still does not hold. A new product line, a seasonal rise or a manager who started coaching the same staff that month could explain the same jump. The sum also counts revenue, while the business keeps only the margin on each extra sale.
- It credits everything to the training, because that is easier than testing it.
- It values revenue instead of profit, which overstates the benefit before any other error.
- It shows no assumption, so a reader cannot tell what would change the answer.
Reports repeat this version because it is fast, not because anyone sets out to mislead. The fix is not a better formula but one extra step before the money: isolating the training’s effect.
How Do You Isolate What the Training Caused?
Isolation separates the training’s share of a result from everything else that moved, and it is the step that makes an ROI claim defensible. Application and business impact have to be observed first, following the Kirkpatrick levels of behavior and results, before there is anything to isolate. The Phillips methodology lists ten strategies and requires at least one before a result reaches the formula, and three cover most programs.
1. Comparison Groups
A comparison group removes most of the guesswork. Two similar groups, one trained and one not, show the training’s effect in the gap between their results.
A true control group is rarely possible at work, but a second store, shift or team trained a month later can serve as one. It works only when the two groups match on role, management layer and season, and it stops working the moment they differ in an obvious way.
2. Trend Lines
Where there is history but no comparison group, a trend line is the next option. Take the measure’s past performance, extend it forward as if nothing had changed, then compare it with the real result.
If daily sales were already rising 1% a week before training started, that rise would have continued anyway. Only the gap above the projected line belongs to the training, which is usually smaller than a raw before-and-after difference.
3. Participant Estimates, Adjusted for Confidence
Where neither of the first two is practical, ask participants or their supervisors to estimate how much of the change the training caused. They list the possible causes first, then give the training’s share.
Each estimate is then discounted for its own uncertainty. People state their confidence on a scale of 0-100%, and the estimate is multiplied by that figure, so $10,000 held with 60% confidence counts as $6,000. The discount is what separates a reported estimate from a hopeful guess.
Converting the Isolated Effect Into Money
Once the training’s share is isolated, it needs a money value so it can be set against the cost. The Phillips paper names three routes that are typical, and the best one is usually whichever already exists in your organization.
1. Find a Standard Value
Finance may already put a price on the measure. A repeat contact, a complaint, a lost customer or a day of absence often carries an agreed cost, developed for some other purpose.
Use that figure rather than inventing a new one. A number finance already trusts survives a review far better than one built for the report.
2. Ask an Expert
Where no standard value exists, ask someone who works with the measure every day, such as the analyst who produces the report or a supervisor. Their estimate rests on a real reference point, which beats a guess from outside the work.
3. Use the Participant’s Own Estimate
Where the people who experienced the change can price it themselves, ask them directly. Apply the same confidence discount you used for isolation, so the estimate carries its uncertainty into the total.
What Belongs on the Cost Side?
Costs go missing almost as often as benefits get inflated, through omission rather than optimism. The Phillips standard is that costs are fully loaded, whether or not anything was invoiced.
Trainer or vendor fees are the easy part, and nearly every report gets them right. What goes missing is usually on this list:
- Wages for time off the job, priced at a loaded hourly rate rather than base salary.
- Travel and venue, for anything the program needed beyond a normal workday.
- Administrative time, for scheduling, coordinating and chasing attendance.
- The evaluation itself, including the hours spent on this calculation.
A loaded wage adds the taxes, benefits and overhead an organization carries for each hour, so it sits well above base pay. Priced that way, participants’ time is often one of the largest costs in the program.
Count each cost once. A participant’s hours belong in the program cost, and they should not appear again inside the benefit as lost productivity. The same gap shows up in coaching ROI in practice, where the coach’s and the client’s time are the costs most often left out.
A Worked Example, Start to Finish
The example below is constructed for illustration, with every step kept arithmetically consistent. It uses a business result, first-contact resolution, rather than a completion count.
Suppose 40 support reps complete a half-day call-handling program, and first-contact resolution rises from 62% to 71% over six weeks. The team also rolled out a new ticketing tool in the same weeks, so supervisors credit half of the gain to training, with 70% confidence. Nine points multiplied by 50% and by 70% gives 3.15 points, which rounds down to 3.
| Step | Working | Result |
|---|---|---|
| Raw gain in first-contact resolution | 71% minus 62% | 9 points |
| Share isolated to training | 9 × 50% × 70%, rounded down | 3 points |
| Standard value per point, set by finance from repeat-contact cost | Given | $1,540 a month |
| Monthly benefit | 3 × $1,540 | $4,620 |
| Benefit for the first year only | $4,620 × 12 | $55,440 |
| Fully loaded cost | Fees $6,000, materials $800, rep time $5,600 (40 reps × 4 hours × $35 loaded rate), administration and evaluation $600 | $13,000 |
| Net benefit | $55,440 minus $13,000 | $42,440 |
| ROI | $42,440 ÷ $13,000 × 100 | 326% |
Skipping isolation would value all nine points, $13,860 a month or $166,320 a year, and show an ROI of about 1,180%. The isolated figure is far smaller, and every discount inside it has a named reason that a finance team can check.
Put the isolation share and the confidence figure in the report next to the final percentage, not in an appendix. A number without its assumption invites exactly the question it was built to survive.
What an Honest ROI Number Still Admits
An isolated, conservative ROI is still an estimate, and the Phillips guiding principles say so in how they handle uncertainty. Three of them are worth stating out loud whenever you present a figure:
- Assume no gain without data. Where no improvement data exists for a group or a source, the standard assumes little or no improvement.
- Leave out extremes. Extreme data items and unsupported claims are dropped from the calculation, whatever they would add to the total.
- Count one year. A short program’s benefit is counted for the first year only, however tempting a longer projection looks.
The isolation share is a judgment, not a measurement, so name who made it and why. “Supervisors credited half of the gain to training because the ticketing tool launched the same month” turns a private guess into a position someone can discuss.
A result this quarter also promises nothing about the next one, because the program and the baseline both change, so treat each calculation as its own case. Comparing the Phillips approach with Kirkpatrick, CIRO or Brinkerhoff is a separate choice, set out in training evaluation models compared. You can run an honest number inside whichever framework you already use.
Training budgets often sit alongside coaching and other programs in the same review, and the wider choice of software for coaches and L&D teams decides where these records live. Keep the baseline, the isolation note and the cost lines together, so the next calculation starts from data rather than memory.
Conclusion
Isolate before you convert, and convert before you divide. Skipping the first step does not make the arithmetic wrong, but it leaves the question a finance team cares about most unanswered.
The conservative figure will be smaller than the quick one, and it is the one that holds up when the budget is decided. Simply.Coach keeps program baselines, outcomes and costs in one place for L&D and HR teams, so the evidence behind the number is ready before anyone asks.
Key Takeaways
- Isolate before you convert. A raw gain is not an ROI input until something separates the training’s share from everything else that changed.
- Discount every estimate. Multiply each estimate by the confidence behind it, or the number carries no record of its own uncertainty.
- Value profit, not revenue. A sales gain is worth the margin the business keeps, not the price on the till.
- Load the full cost. Time off the job, administration and evaluation belong in the total alongside the invoice.
- Count each hour once. Time already in the cost should not reappear as lost productivity in the benefit.
- Round down when unsure. The conservative choice is the one a finance team can accept.
- Show the assumption. Print the isolation share beside the final percentage.
If the hardest part of a training review is finding the baseline, the costs and the outcomes in one place, the number will always be weaker than the work. Simply.Coach’s HR and L&D coaching software gives teams reporting a sponsor can read, with program records that hold up to a second question.
Frequently Asked Questions (FAQs)
Which is the best training outcome measure?
No single measure suits every program; the right one follows from what the training was meant to change, such as error rates, sales or resolution times. What matters more is isolating the training's own effect on that measure before you report it, which is the step most reports skip.
How do you measure the effectiveness of a program?
Measure in layers: how participants reacted, what they learned, whether they applied it at work and what changed in the business. For a money figure, isolate the program's share of that change, convert it to money and compare it with the fully loaded cost.
What are KPIs for training?
Common ones include completion and assessment scores, the share of participants applying the skill on the job, and business measures such as error rates, output or first-contact resolution. Choose the business measure the program was meant to move, and agree it before the program starts.
What are the 5 levels of training evaluation?
In the Phillips extension of Kirkpatrick's model, the five levels are reaction, learning, application, business impact and return on investment. Each level builds on the one below, so an ROI figure needs evidence of application and impact before the money calculation starts.
What is the formula for calculating training effectiveness?
For ROI, the formula is net benefit divided by fully loaded cost, multiplied by 100. Net benefit is the isolated, money value of the training's effect minus everything the program cost. Both halves of that sentence decide whether the result holds up.
What is a good training ROI to aim for?
There is no universal benchmark, and a percentage quoted without its isolation method cannot be compared with yours. Judge a result on how it was built: whether it isolates the effect, loads the full cost and counts only the first year. A modest, defensible figure beats a large one.
Can you calculate training ROI without a control group?
Yes. A comparison group is one of several ways to isolate the effect, and trend lines or participant estimates discounted for confidence are more common. Keep a true comparison group for the most expensive or most contested programs, where the extra analysis is worth its cost.
Does training ROI include the trainer's own time?
It should. Price an internal trainer's hours at a loaded rate and use the invoiced fee for an external one. Leaving the trainer's or coach's time out is one of the most common reasons a reported cost understates the real one.