7 Proven Ways to Measure Training ROI and Why It Matters

The seven ways to measure training ROI are: the Kirkpatrick four levels, the Phillips ROI formula, a control group comparison, triangulated attribution, tracking a business KPI the company already watches, comparing retention between trained and untrained staff, and forecasting ROI before the programme starts.

Most teams should start with the last one. It is the cheapest, and it makes all the others easier.

The Question Every L&D Team Dreads

A CFO opens the training line in the budget and asks a simple question. We spent this much on leadership development last year. What did we get back?

The L&D lead has satisfaction scores, completion rates and good feedback. None of it answers the question.

This is not a small problem. Research suggests only about 8 percent of L&D leaders can show the business impact of training in financial terms. Around 65 percent of L&D teams never get past measuring whether people learned something, and never reach the level where they measure whether anything changed in the business.

Meanwhile companies spend roughly 1,200 to 1,600 US dollars per employee per year on learning. That is a large number with a weak story attached to it, which is why training budgets are usually one of the first things reviewed when costs get cut.

Why Most Measurement Fails

Three reasons, and they are all fixable.

Nobody wrote down the starting point. You cannot show improvement if you never recorded where you began. This is the single most common failure, and it happens because measurement gets planned after the programme rather than before it.

The wrong things get counted. Completion rates and satisfaction scores are easy to collect and tell you almost nothing. High completion does not mean behaviour changed.

Measuring costs more than it is worth. One analysis put the labour cost of properly evaluating a single cohort at around 15,000 US dollars, which can be more than the training platform itself. If measurement costs more than the answer is worth, nobody will do it.

The seven methods below are ordered roughly by cost and effort, from simplest to hardest.

Quick Comparison

1. Forecast ROI Before the Programme Starts

Most people think of measurement as something you do after training. That is backwards, and it is why so much measurement fails. The strongest thing you can do costs one afternoon and happens before anything is booked.

What it is

Instead of measuring after, you model the expected return before you ask for budget. You pick the number you expect to move, estimate how much it will move, put a money value on that movement, and compare it to what the programme costs.

What it looks like in practice

Say you are running a service programme for a 40 person contact centre. Your average handling time is 9 minutes. You expect the programme to bring it to 8 minutes, which is about 11 percent. Your team handles 4,000 calls a month. One minute saved per call is roughly 67 hours of staff time a month. At an average loaded cost of 25 dollars an hour, that is about 1,675 dollars a month, or 20,100 a year. The programme costs 15,000 dollars including participant time. You now walk into the budget meeting with a forecast instead of a hope. And you have written down that 9 minutes is where you started, which is the part that makes everything else possible later.

Why this changes the conversation

Leadership is not usually against training. They are against spending on something with no visible outcome. A forecast turns an expense into a proposal. It also protects you. If you said 11 percent and delivered 9 percent, that is a good result you can defend. If you said nothing and delivered 9 percent, you have nothing to point at.

What to watch out for

Do not be optimistic. Pick a number you are confident about beating. The temptation is to forecast something impressive to get the budget approved. Resist it. A forecast you exceed builds trust for the next request. A forecast you miss makes the next request harder, even if the programme genuinely worked.

2. Track a Business KPI the Company Already Watches

This is the cheapest ongoing method, and for most programmes it is the only one you actually need.

What it is

Pick one number the business already tracks and already cares about. Record where it stands before the programme. Check it again after. That is the whole method.

Choosing the right number

The number should be close enough to the training that the connection is obvious to anyone.

The left column moves because of the training. The right column moves because of a hundred things, and the moment you claim credit for it, someone in the room will point that out.

Why it works better than a custom metric

You are not asking leadership to accept a new measurement. They already look at this number every month, they already trust it, and they already know what good looks like. That saves you an entire argument. Instead of defending your metric, you are just showing movement in theirs.

What to watch out for

Set the baseline before the programme, not after. This sounds obvious and it is the single most common failure in training measurement. Also record what else was happening. If you ran the training during a quiet quarter, or right after a new system went live, note it. Someone will ask, and having the answer ready is better than being caught out.

3. The Kirkpatrick Four Levels

This is the oldest and most widely used training evaluation framework. It is worth knowing properly, because most organisations use the name while only doing a quarter of the work.

The four levels

Where most organisations stop

At Level 2. They run a happy sheet, maybe a quiz, and file the results. That tells you people enjoyed a session and remembered some content on the day. It tells you nothing about whether anyone works differently now, which is the only reason you paid for the training.

Real evidence begins at Level 3.

Making Level 3 actually work

Level 3 fails when it becomes a form-filling exercise. Managers get asked to rate whether their team member has improved, they tick the middle box, and everyone moves on. It works when managers know beforehand what to look for. Give them three specific things to watch for during the 90 days, in plain language.

For a service programme that might be: does this person now acknowledge the issue before offering a fix, do they explain what happens next without being asked, and do they check the customer is satisfied before closing.

Three specific behaviours. Not a rating scale.

What to watch out for

Do not skip Level 1 and 2 entirely just because they are weak on their own. They are cheap, and they help you diagnose. If people learned nothing at Level 2, you know the problem is the content. If they learned it but Level 3 shows no change, the problem is back at work, not in the room.

4. Triangulated Attribution

Here is the hardest question in training measurement. The number moved. How much of that was the training, and how much was everything else? There is no perfect answer. This method gets you a defensible one.

What it is

Ask two groups the same question, separately, then average their answers. Ask participants: of the improvement you have seen in this area, what percentage do you credit to the training? Then ask their managers, independently, the same question about the same people.

Why averaging works

Participants tend to overstate. They invested time and want it to have mattered. Managers tend to understate. They see other factors, including their own coaching, and they were not in the room. The two biases pull in opposite directions. Averaging them lands closer to reality than either one alone, and more importantly, it is a method you can explain in a meeting without anyone accusing you of picking a convenient number.

A worked example

Complaints dropped from 120 a month to 90. That is 30 fewer complaints. Participants say 70 percent of that was the training. Managers say 40 percent. The average is 55 percent. So you claim credit for roughly 16 of the 30 complaints, not all 30. Claiming 16 makes you credible. Claiming 30 makes everything else you say suspect.

What to watch out for

Timing. Run this around 90 days after the programme. Any earlier and behaviour has not settled. Any later and memory has faded to the point where the answers are guesses. Also keep the two surveys genuinely separate. If managers and participants discuss it first, you get one answer twice, not two independent ones.

5. Compare Retention Between Trained and Untrained Staff

This is the method most L&D teams overlook, and in a tight labour market it is often the strongest argument available.

What it is

Compare how long people stay, between those who completed a programme and similar colleagues who did not.

Why it is powerful

Turnover already has a money value attached, and finance usually already knows it. Recruitment fees, notice periods, onboarding time, and the months of reduced productivity while a new person learns the job.

That means you do not have to convince anyone the saving is real. You just have to show the difference in turnover and multiply.

A worked example

Twenty five managers completed a leadership programme. Over the following year, two left. That is 8 percent. Across comparable managers who did not attend, turnover was 20 percent. That is a 12 percentage point difference, which is roughly three people who stayed and might otherwise have gone. If your organisation costs replacement at 50,000 dollars per manager, that is 150,000 dollars of avoided cost against a programme that might have cost 40,000.

What to watch out for

Compare like with like. If the trained group were high performers hand-picked for development, they were probably more likely to stay anyway. Match on role, seniority and tenure as closely as you can, and say openly where the match is imperfect.

Being upfront about the limitation is what makes the rest of the number believable.

6. The Phillips ROI Formula

This is the one that produces the percentage a CFO recognises. It builds on Kirkpatrick by adding a fifth level that converts business results into money.

The formula

ROI = ((Benefits minus Costs) divided by Costs) times 100

Counting the costs honestly

This is where most calculations fall apart. Costs are not just the invoice. Include the programme fee, participant time at their loaded hourly cost, facilitator time if internal, travel and venue, materials and platform costs, and the cost of doing the measurement itself. Participant time is usually the biggest line and the one people leave out. Twenty people for two days is 320 hours. At 40 dollars an hour that is 12,800 dollars, which may well exceed the programme fee. Leave it out and a CFO will find it. Include it and you look like someone who understands the business.

A worked example

Using the retention numbers above: benefits of 150,000, costs of 40,000. 150,000 minus 40,000 is 110,000. Divided by 40,000 is 2.75. Times 100 is 275 percent. But apply the attribution percentage from method 4 first. If attribution is 55 percent, your benefit is 82,500 rather than 150,000. 82,500 minus 40,000 is 42,500. Divided by 40,000 is 1.06. Times 100 is 106 percent. A defensible 106 percent beats an inflated 275 percent every time.

What good looks like

Published benchmarks generally place a healthy training ROI between 25 and 300 percent, meaning roughly 1.25 to 4 dollars back for every dollar spent. If your number lands far above that range, something in the calculation is too generous, and the meeting will not go the way you hope.

What to watch out for

Do not run this on every programme. It takes real work and it needs Level 4 results plus an attribution figure. Save it for the programmes big enough to justify the effort, or for the one you know you will be asked to defend.

7. The Control Group

The strongest proof available, and the hardest to arrange.

What it is

Train one group. Do not train a similar group. Compare the two over the same period.

Why it settles the argument

Every other method leaves room for the same objection: how do you know it was the training and not the market, the new system, or the season?A control group answers it. Both groups faced the same quarter, the same market and the same conditions. Only one had the training. If only one improved, the case is hard to dispute.

A worked example

Two sales regions of similar size and maturity. Region A gets the programme in January. Region B does not. Over six months, Region A's win rate goes from 22 percent to 28 percent. Region B goes from 23 percent to 24 percent. The market moved everyone by about one point. The training accounts for the other five. No attribution survey needed. The comparison does the work.

Making it politically workable

Somebody has to be told they are not getting the training yet, and that can create resentment if handled badly. Frame it as a phased rollout, which is usually what it genuinely is. Region B goes second, in the next cycle. Tell them that clearly and give them a date. Most organisations run phased rollouts anyway for budget reasons. The only extra step is deciding to measure the gap between the phases.

What to watch out for

The groups have to be genuinely comparable. Different markets, different managers or different product mixes will produce a difference that has nothing to do with training, and someone will spot it.

Three Numbers to Stop Reporting

These get reported because they are easy to collect, and they cost you credibility.

Completion rates. This tells you people attended. If everyone completed the training and nothing changed, that is not success.

Satisfaction scores. These measure whether a session was pleasant. Enjoyment and behaviour change are barely related. A session people enjoyed and forgot scores well.

Hours delivered. This is a measure of activity, not outcome. Reporting it to a CFO invites the obvious question of whether fewer hours would have done the same job.

Replace all three with one sentence: this number moved from here to here, and here is what we think we contributed.

What to Do This Quarter

If you are starting from nothing, this is the order that works.

Before your next programme, write down one number you expect to move, where it is today, and by how much you expect it to change. That is method 1, and it takes an afternoon.

At 90 days, check that number and run the triangulated attribution survey. That is methods 2 and 4.

At six months, if the number moved, build the Phillips calculation. That is method 6, and by then you have everything you need for it.

Do not try to do all seven on your first programme. Pick one number and measure it properly.

Why Measurement Changes the Training Itself

There is a side effect worth mentioning.

Once you have to name a number before the programme starts, the design of the programme changes. Vague objectives like better leadership do not survive the question of what number will move. You are forced to get specific about the outcome, and specific programmes work better than broad ones.

We set the success criteria before design begins across our corporate training in Singapore work for exactly this reason. A measure chosen after delivery will always be chosen to flatter the result.

It also surfaces something uncomfortable but useful. Sometimes the number does not move because the problem was never a training problem. If people knew how to do the thing but were not allowed to, or were measured on something else, no programme was ever going to fix that. We wrote about how to tell the difference in our comparison of leadership training and broader corporate training.

And if the number moves at first and then drifts back, the issue is usually reinforcement rather than measurement, which we covered in why employee training does not stick.

Frequently Asked Questions

How do you calculate training ROI?

Use the formula ROI equals benefits minus costs, divided by costs, times 100. Benefits are the money value of what improved. Costs must include everything, meaning programme fees, participant time, travel, facilitator time and the cost of measuring. Leaving out participant time is the most common mistake and the easiest for a CFO to spot.

What is a good training ROI?

Published benchmarks generally place a healthy range between 25 and 300 percent, meaning roughly 1.25 to 4 dollars returned for every dollar spent. Figures far above that usually mean something in the calculation is too generous. Be conservative, because a number you can defend is worth more than a big one you cannot.

What is the Kirkpatrick model?

A four level framework for evaluating training. Level 1 measures reaction, meaning whether people liked it. Level 2 measures learning, meaning whether they retained anything. Level 3 measures behaviour, meaning whether they work differently around 90 days later. Level 4 measures results, meaning whether business numbers moved. Most organisations stop at Levels 1 and 2, which is why so few can demonstrate business impact.

Why can most L&D teams not prove training ROI?

Usually because nobody recorded the starting point before the programme began. Without a baseline there is nothing to compare against. The other common causes are measuring easy things like completion and satisfaction instead of behaviour and results, and measurement that costs more effort than the answer is worth.

How long after training should you measure ROI?

Behaviour change is usually visible around 90 days. Business results generally take two to three quarters, because they depend on work that happened after the change. Measuring immediately after a session only captures enthusiasm. Waiting a full year means the window to correct anything has closed.

What should you stop reporting to leadership?

Completion rates, satisfaction scores and hours delivered. All three measure activity rather than outcome. Replace them with a single statement showing one business number, where it started, where it is now, and what portion of that change you believe the training contributed.