Your last training program scored 4.6 out of 5. Participants loved it. Then the CFO asks one question:
what did it return?
Most HR and L&D teams cannot answer. According to ATD's 2026 State of the Industry report, only 13% of organisations measure ROI across all their learning programs, and more than half do not measure it at all.
That gap is becoming expensive. Budgets are tightening: SHRM's 2026 benchmarking found median L&D spend per employee fell 28% in a year, even as training hours stayed flat. When money is scarce, the programs that survive are the ones that can show a return in the CFO's language.
I have sat on both sides of that table. As a CFO, I have reviewed budgets line by line. As a trainer, I have delivered programs to more than 17,000 professionals. This article gives you the method I use to bridge the two: six steps, a handful of formulas and a worked example you can adapt to your next program.
One promise before we start. The method is deliberately conservative. A modest ROI that the CFO believes is worth far more than a spectacular one that nobody trusts.
Why most training ROI numbers fail
When training ROI is reported at all, it usually breaks in one of three ways.
- It measures happiness, not results. Feedback scores and completion rates show that people attended and enjoyed it. They say nothing about what changed at work.
- It counts only the invoice. The trainer's fee is the visible cost. The largest cost is usually invisible: the salary of every participant for every hour they were in the room.
- It claims all the credit. Sales rose 12% after a sales training. So did the market, the price list and the incentive scheme. A CFO will discount any number that ignores this, and often the whole report with it.
The fix is a ladder that most L&D professionals already know, with one extra rung that finance teams care about.
The five levels, and where most companies stop
Donald Kirkpatrick's model evaluates training at four levels. Jack Phillips added a fifth, which converts results into money and compares them with cost.
| Level | Question it answers | Typical evidence |
|---|---|---|
| 1. Reaction | Did they like it? | Feedback form |
| 2. Learning | Did they learn it? | Test, assessment, demo |
| 3. Application | Are they using it at work? | Manager check-ins, system data at 30, 60 and 90 days |
| 4. Impact | Did a business metric move? | Time, errors, sales, attrition, compliance findings |
| 5. ROI | Was the money well spent? | Monetary benefit compared with full cost |
Most organisations stop at Level 1. The method below gets you to Level 5, credibly.
The six-step method
Step 1: Choose the business metric before you design the program
ROI cannot be calculated after the fact if nobody recorded the starting point. Before the program is designed, agree with the business head on one or two metrics it should move, and record today's value.
Good metrics are already tracked somewhere, can be converted to money, and change within three to six months. For example:
- Hours spent on a recurring task (month-end close, report preparation, reconciliations)
- Error or rework rate (invoice errors, rejected claims, audit findings)
- Output per person (calls handled, proposals sent, cases closed)
- Sales conversion or average deal size
- Attrition in the trained group
If no one can name the metric, the program may be worth running for other reasons, but its ROI cannot be measured. Say so upfront.
Step 2: Count the full cost, not just the invoice
Finance teams call this the "fully loaded" cost. Leave items out and the ROI looks better on paper, but it will not survive the CFO's review.
| Cost item | How to calculate it |
|---|---|
| Design and delivery | Trainer or consultant fees, content development |
| Participant time | Hours in training × hourly cost to company (annual CTC ÷ about 2,000 working hours) |
| Tools and licences | Software or platform seats bought for the program |
| Logistics | Venue, travel, food, printed material |
| Administration | Internal HR or L&D time to organise and evaluate |
Participant time is the line most often missed, and frequently the largest. Forty people at ₹12 lakh CTC spending 16 hours in training cost about ₹3.8 lakh in salary alone.
Step 3: Track application at 30, 60 and 90 days
A skill learned but not used returns nothing. Before you measure any business result, check how many participants are actually using what they learned.
Use evidence, not opinion. Ask managers whether each person is applying the skill, check system data where it exists, and review a sample of work. Record the share of participants applying it at 90 days. Only that group's results count toward the benefit.
Step 4: Convert the change into money
Most training benefits fall into one of four routes. Use the most conservative value that finance will accept.
| What improved | How to value it | Watch out for |
|---|---|---|
| Time saved | Hours saved × hourly cost to company | Count it only if the time goes to useful work, not idle capacity |
| Fewer errors or less rework | Errors avoided × cost per error (write-offs, penalties, rework hours) | Use actual cost per error from your records, not an estimate |
| More output or sales | Extra output × contribution margin | Use margin, never revenue |
| Lower attrition | Leavers avoided × replacement cost | Replacement benchmarks run from 50% to 200% of salary; use the low end |
Two rules keep this credible. Annualise benefits over 12 months at most for a short program; don't project three years of gains from a two-day workshop. And keep benefits you cannot value honestly, such as morale or confidence, in a separate "intangible" list rather than forcing a number onto them.
Step 5: Isolate the effect of the training
This is the step that makes a CFO trust the number. Other things changed during the same period, so you must separate the training's share. There are three practical ways to do it, from strongest to simplest.
- Control group. Train one team and compare it with a similar team that was not trained. The difference is the training effect.
- Trend line. Project where the metric was heading before the program. Count only the improvement above that trend.
- Estimates adjusted for confidence. Ask participants and their managers what share of the improvement came from the training, and how confident they are in that estimate. Multiply the improvement by both. If someone says 70% of a gain came from training and they are 75% confident, you count 52.5% of it.
The third method is the most common in practice, because control groups are rarely available. It is also the most transparent: every discount is visible to the reader.
Step 6: Calculate ROI, benefit-cost ratio and payback
Four numbers, three of them answering a question a CFO will ask:
- Net benefit = Monetary benefits − Full cost
- ROI (%) = Net benefit ÷ Full cost × 100. Did we make more than we spent?
- Benefit-cost ratio = Monetary benefits ÷ Full cost. How much came back for every rupee?
- Payback (months) = Full cost ÷ Monthly benefit. How soon do we recover the money?
An ROI of 0% means you broke even. Anything above it is a return; anything below it is a loss worth learning from.
A worked example: AI upskilling for a finance team
A mid-sized company trains its 40-person finance and operations team to use AI tools in daily work: drafting reconciliations, summarising vendor statements, preparing first-cut reports. The agreed metrics are hours spent on recurring reporting tasks, and payment errors caught before release.
Full cost
| Item | Calculation | Amount |
|---|---|---|
| Design and delivery | Trainer fees and content | ₹6,00,000 |
| Participant time | 40 people × 16 hours × ₹600 an hour | ₹3,84,000 |
| AI tool licences | 40 seats × ₹2,000 a month × 3 months | ₹2,40,000 |
| Logistics and administration | Venue, material, HR time | ₹76,000 |
| Total | ₹13,00,000 |
Application at 90 days: 32 of the 40 participants are using the tools on recurring tasks every week. Only these 32 are counted.
Benefits over 12 months, before isolation
- Time saved: 32 people × 3 hours a week × 46 working weeks = 4,416 hours × ₹600 = ₹26,49,600
- Duplicate or incorrect payments caught before release: ₹3,50,000
- Total claimed: ₹29,99,600
After isolation
- Time saved: managers attribute 70% of the time saving to the training, with 75% confidence. ₹26,49,600 × 70% × 75% = ₹13,91,040
- Payment errors: 60% attributed to the training. ₹3,50,000 × 60% = ₹2,10,000
- Total credible benefit: ₹16,01,040
The result
| Measure | Before isolation | After isolation |
|---|---|---|
| Net benefit | ₹16,99,600 | ₹3,01,040 |
| ROI | 131% | 23% |
| Benefit-cost ratio | 2.31 | 1.23 |
| Payback | about 5 months | about 10 months |
Look at the gap between the two columns. A team eager to impress would report 131%. A CFO would take that number apart in one meeting, and with it, the credibility of every future request.
The 23% is the number to report. It is positive, it survives scrutiny, and it was built from discounts the reader can see. The benefits we could not value honestly, such as faster responses to auditors and higher confidence with new tools, go in a separate list of intangibles.
Report it on one page
Leadership will not read a 20-slide evaluation deck. Put the whole story on one page, in this order:
- The business problem and the metric agreed upfront, with its starting value
- What was delivered: who was trained, for how long, at what full cost
- Application: the share of participants using the skill at 90 days
- Impact: the change in the metric, before and after isolation
- ROI, benefit-cost ratio and payback, with every assumption listed underneath
- Intangible benefits in a short list
- What we will change next time
Listing your assumptions is not a weakness. It is what makes the number credible, and it invites the CFO to adjust an assumption rather than reject the whole result.
Do you need ROI for every program?
No. Full ROI analysis takes effort, and some training, such as mandatory compliance courses or induction, exists for reasons other than financial return. Measure Levels 1 to 3 for everything. Reserve the full Level 5 analysis for programs that are expensive, strategic, or likely to be repeated across the organisation. Those are the programs where a credible number changes budget decisions.
Frequently asked questions
What is a good ROI for a training program?
Any positive ROI means the program returned more than it cost, after accounting for everyone's time. What matters more than the size is credibility: a conservative 20% to 30% with visible assumptions is far more persuasive than an unsupported 300%.
How soon after training can ROI be measured?
Check application at 30, 60 and 90 days. Most business metrics need three to six months to show a reliable change. Measuring the day after the program only tells you how participants felt.
What if we did not record a baseline?
Use historical data from the months before the program, such as past reports, system logs or payroll records. If none exist, measure this program's Levels 1 to 3 properly and set a baseline for the next one.
How do I value soft skills such as leadership or communication training?
Link them to a metric they should influence, such as team attrition, escalations or time to close a deal, and isolate conservatively. Benefits that cannot be valued honestly belong in the intangibles list.
Should participant salaries really count as a cost?
Yes. The time employees spend in training is time not spent on their normal work. Leaving it out is the most common reason a training ROI fails a finance review.
Want a credible ROI on your next program?
Grevx designs corporate training programs in finance, AI and business skills with measurement built in from day one: an agreed metric, a baseline, a 90-day application check and a one-page ROI report for leadership. We run programs for teams in India, the UAE and beyond.
If you are planning next year's training calendar, talk to us. Bring one program you need to justify, and we will show you how to measure it.
Need help applying this insight?
Our consulting team helps businesses translate strategy into practical action.
Talk to our team →