On this page
The formula for learning and development ROI takes about a minute to learn. Net benefits divided by costs, times a hundred. Every article on this subject gives it to you, and none of them is wrong.
It is also not the reason your ROI number gets challenged in the budget meeting. That happens at the next question, the one about how you know the training caused the improvement rather than the new manager, the pricing change, or the quarter simply being better.
L&D ROI expresses the financial return on training as a percentage: net benefits divided by fully loaded costs. It sits at Level 5 of the Phillips methodology, above Kirkpatrick's four.
The arithmetic is trivial. What makes a number credible is isolating training's contribution from everything else happening at the same time, and that requires planning before the programme runs.
- Cost the problem before you design the solution
- Capture a baseline, because it cannot be retrofitted
- Choose an isolation method while you can still design for it
- Include participant time in the cost side
- Report a range with your method stated, not a single confident number
Capability baselines captured during rollout rather than reconstructed afterwards, so the before figure exists when you need it.
What Is Learning and Development ROI?
L&D ROI is the financial return an organisation gets from money spent on training, expressed as a percentage of the cost. It is a specific calculation rather than a general claim that training was worthwhile.
It also has a specific home in evaluation theory. Donald Kirkpatrick's four-level model measures reaction, learning, behaviour and results. Jack Phillips added a fifth level that takes those business results and converts them into monetary value, then sets that value against the fully loaded cost of the programme. Level 5 is where ROI lives, and it is the only level that produces a number a finance director recognises.
Two things follow from that placement, and both get missed.
Level 5 depends on Level 4. You cannot convert business results to money if you never measured business results. An organisation reporting completion rates and satisfaction scores is operating at Levels 1 and 2 and has no path to an ROI number, however good its arithmetic.
And ROI is a percentage, not a proof. It says the numbers worked out. Whether the numbers were the right ones, and whether the improvement was caused by the training, are separate questions the percentage does not answer.
Simple: ROI is Level 5 of a five-level chain. Skipping to it from Level 1 produces a number with nothing underneath it.
Key Evaluation Frameworks
ROI does not stand alone. It is the top of a stack, and knowing which stack you are in changes what you are being asked for.
Kirkpatrick is the foundation, published by Donald Kirkpatrick from 1959. Four levels: reaction, learning, behaviour, results. It stops at business results and deliberately does not monetise them.
Phillips adds the fifth. Jack Phillips took Kirkpatrick's Level 4 outputs, converted them to monetary value, set them against fully loaded cost, and supplied the isolation techniques that make the conversion defensible. When someone asks for training ROI, this is the framework they are asking for whether or not they name it.
Five others get used in practice, each answering a different question: CIRO for context and input alongside outcomes, Kaufman for societal and customer impact beyond the organisation, Anderson for alignment with strategic priorities, Brinkerhoff's Success Case Method for what worked and why rather than what averaged out, and Thalheimer's LTEM for whether learning actually transferred to the job.
Two practical notes. The frameworks are layered rather than competing, so Phillips assumes Kirkpatrick underneath it and LTEM reacts to both. And you rarely need more than one: pick the framework that answers the question your stakeholder actually asked. For the full comparison and how to choose between them, training evaluation models covers all seven properly.
How to Calculate L&D ROI Step-by-Step
Here it is, so nobody has to go looking:
ROI (%) = (Net programme benefits ÷ programme costs) × 100
Net benefits are the monetary value of the improvement minus the cost of the programme. A result of 100% means you got your money back plus the same again.
Two things about the inputs are worth more attention than the formula itself.
- Costs must be fully loaded. The licence fee and the facilitator invoice are the visible part. The larger number is almost always participant time: two hundred people off the job for a day is two hundred days of salary plus whatever they were not doing while they were in the room. Leave it out and the ROI looks excellent right up until finance adds it back and asks why you did not.
- Benefits need converting, and some convert badly. Revenue, cost reduction and time saved are straightforward. Fewer escalations, lower rework and reduced attrition need a proxy value, which means an assumption that has to be stated. Anything you cannot value credibly belongs in an intangibles list rather than forced into the numerator.
That is the arithmetic, and it is the part every article covers and almost nobody gets wrong. The procedure around it is where the work sits.
- Cost the problem before designing the solution Put a number on what the problem is costing now: error rates, rework hours, attrition replacement cost, lost sales, time to competence. That figure becomes the reference point for everything after it, and it is far easier to establish while the problem is live than to reconstruct a year later. A training needs analysis that stops at what people need to learn has skipped the number that makes the programme defensible.
- Capture the baseline Record the metric you intend to move, for the affected population, before anything starts. Record it for a comparable population too if one exists. This is a week's work at the beginning and impossible at the end, and its absence is the single most common reason an ROI figure cannot be produced.
- Choose the isolation method now, not later Decide how you will separate the training's contribution while you can still design for it. A phased rollout has to be sequenced deliberately. Trend-line projection needs enough history to establish a trend. Estimation needs the questions written before people forget what happened. Deciding afterwards leaves you with whichever method your data happens to permit, which is usually the weakest. The four options are in the next section.
- Agree the measure with whoever will judge it Get the budget holder to confirm, in writing, what counts as success and over what horizon. One conversation, and it prevents the review where the number you spent a year producing turns out not to be the one they cared about.
- Run the programme and wait Behaviour change shows up in months, not weeks. Measuring at thirty days captures enthusiasm; ninety days captures early application; six to twelve months captures whether it held. Report at more than one horizon if you can, because a figure that improves over time is more convincing than a single snapshot.
- Isolate, convert, calculate, and state your method Apply the isolation technique you chose, convert the isolated share to monetary value, subtract fully loaded costs, divide, multiply by a hundred. Then write down how you isolated and what you assumed. The method is what makes the number survive the meeting.
Isolating Training's Contribution
Here is the question that ends most ROI presentations. Sales went up, but we also launched a campaign and hired two people. How do you know it was the training?
Phillips identified this as the methodological crux, and it is the step the published literature on this topic most often skips. Without it you have a correlation and a hopeful inference. With it you have a defensible share.
Four techniques, in descending order of rigour and ascending order of feasibility.
- Control or comparison group. Methodologically the strongest and rarely available, because withholding training from half a team is a conversation nobody wants to have. Where it exists, it is usually because a phased rollout created it by accident.
- Phased or stepped rollout. Train sites or teams on a staggered schedule, so each becomes its own before-and-after case and the later groups act as an interim comparison. This is the practical version of a control group and it costs nothing extra, because most large rollouts are phased anyway. The only requirement is that somebody decides to measure at each phase rather than once at the end, which is a scheduling decision rather than a methodological one.
- Trend-line projection. If the metric was already improving before the training, extend that trend forward and measure only the gap between the projected line and the actual result. Everything below the line was happening anyway. This is the most useful technique for organisations with decent historical data and no ability to hold a group back, and it has a second benefit worth more than the first: it prevents the most common overclaim in L&D reporting, which is taking credit for a trend that was already running before anyone booked a room.
- Participant and manager estimation. Ask both independently what share they attribute to the training, average the two, apply a confidence discount. Weakest of the four, legitimate within the methodology, and only defensible if you label it as an estimate.
Mistakes: Presenting an ROI figure without saying how the isolation was done. The method is the credibility. A 140% return with trend-line projection stated is more persuasive than a 400% return with no method, because the second one invites the question the first one has already answered.
Which Programmes Deserve an ROI Study
Not all of them, and saying so is more credible than pretending otherwise.
Phillips said so himself: the methodology is time-consuming and expensive, and should be applied selectively. A full Level 5 study needs baselines, isolation, monetary conversion and a defensible cost model. Running that on every programme exhausts a small team and produces a stack of numbers nobody reads.
Worth the effort:
- Large, expensive programmes where the spend itself invites scrutiny
- Programmes attached to a metric finance already tracks, where conversion is straightforward
- Anything you intend to expand, since the business case for scaling needs a number
- Programmes where a previous version was cut and you are rebuilding the case
Not worth it:
- Small pilots, where the study costs more than the programme
- Compliance training, where the benefit is avoided risk rather than gained value and the real answer is that the alternative is a penalty
- Anything where the outcome is genuinely long-horizon, such as leadership development, where the attribution chain is too long to isolate honestly
- Programmes that were mandated rather than chosen, where the ROI question is academic
For the rest, Levels 1 to 4 are usually sufficient, and honest. Behaviour changed, the business metric moved, here is the evidence. That is a strong report without a percentage attached.
What to Report When ROI Is Not Available
Most of the time you will not have a defensible ROI figure, and the choice is not between ROI and nothing.
Four measures that survive scrutiny without requiring monetary conversion:
- Movement in assessed capability, by cohort, against a defined role requirement. Concrete, comparable over time, and it answers whether anything changed.
- Time to competence, which converts to cost directly enough that finance can do the conversion themselves if they want to.
- Internal fill rate, the proportion of open roles filled internally, which connects development to a cost the business already feels. This is where career mobility data earns its place in an L&D report.
- Operational metric movement in the trained population, stated alongside what else was happening, which is more honest than an isolated figure and often more persuasive.
The general principle is to report the chain rather than the conclusion. Here is what we measured before, here is what we did, here is what moved, here is what else was going on, and here is the share we think is attributable and why. A finance director will accept a range with reasoning more readily than a single number with none. For the reporting infrastructure this depends on, learning analytics covers the data layer.
Archana Bhaskar, CHRO of Dr. Reddy, India, shares effective tips to assess if the time, money, and resources invested in L&D are reaping results and succeeding in meeting business needs. Check out the full conversation of the L&D expert with Sunita Arora, L&D advisor, and strategist, as she unveils tactful ways for assessing the return on L&D investments.
How Disprz Helps
The recurring obstacle in everything above is the baseline. Almost every organisation that cannot produce an ROI figure is in that position because nobody captured a before, and by the time the question arrives it is too late.
Disprz addresses that at the point of rollout rather than at the point of reporting. Skill mapping defines what each role requires and at what proficiency, and assessment produces a measured capability position per person, which means the before figure exists as a by-product of setting the programme up rather than as a separate exercise somebody has to remember. When the question arrives twelve months later, the baseline is already there.
From there, capability movement is reported by cohort against the defined requirement, which gives the Level 3 and Level 4 evidence the ROI calculation needs underneath it. Where the rollout is phased, comparing trained and untrained cohorts is a query rather than a study, which makes the most practical isolation method available by default. Turo converts existing documentation into role-specific content 80% to 90% faster than a conventional authoring cycle, with human review retained, which matters for the cost side: a large share of programme cost is authoring time, and reducing it changes the denominator as well as the timeline.
The system holds the record, and analytics report skill movement rather than hours consumed, which is the difference between a report that supports an ROI case and one that documents activity. Across 3.5 million learners in 500+ organisations and 25+ countries, the pattern that holds is that programmes with a baseline can answer the question and programmes without one negotiate.
For the broader question of connecting development to business outcomes rather than proving a single programme, talent development covers the programme layer.
If you are being asked for an ROI figure on a programme that has already run without a baseline, the honest answer is that you can produce an estimate with a stated method and a confidence range, and not a measurement. Say that, show the chain, and put the baseline in place for the next one. It is a better position than a confident number that does not survive the second question.
FAQs
What is learning and development ROI?
Learning and development ROI is the financial return on training spend, expressed as net programme benefits divided by fully loaded costs, times one hundred. It sits at Level 5 of the Phillips methodology, which extends Kirkpatrick's four levels by converting business results into monetary value.
How do you calculate training ROI?
Divide net programme benefits by total programme costs and multiply by a hundred. The arithmetic is simple; the difficulty is establishing credible inputs, particularly fully loaded costs including participant time, and a benefits figure that isolates what the training actually caused.
What is the Phillips ROI model?
The Phillips ROI model extends Kirkpatrick's four evaluation levels with a fifth. Levels one to four cover reaction, learning, behaviour and business results; Level 5 converts those results into monetary value and compares them with programme cost. It also specifies techniques for isolating training's contribution.
How do you isolate the effects of training?
Four techniques exist. Control or comparison groups are strongest but rarely feasible. Phased rollouts create comparisons naturally. Trend-line projection measures the gap above a pre-existing trend. Participant and manager estimation is weakest and acceptable when the method and confidence discount are stated.
Why is L&D ROI so hard to measure?
The obstacle is rarely the formula. Most programmes were never designed to be measured: no baseline was captured, no comparison group exists, and no isolation method was chosen. None of those can be added retrospectively, so the calculation becomes impossible after the fact rather than merely difficult.
What should you report if you cannot calculate ROI?
Report movement in assessed capability by cohort, time to competence, internal fill rate, and operational metric movement stated alongside other contributing factors. A range with the method explained is more credible to a finance audience than a single figure with no stated basis.
