Measuring training: four levels and an honest ROI
Short answer
Training is best measured on four levels: participants' reaction, learning, changed behaviour on the job, and results for the business. The higher the level, the more meaningful and the harder the measurement. A return on investment can only be stated honestly where a result is clearly measurable and other influences are accounted for.
Key points
- Satisfaction says little about impact — it is only the first level.
- Behaviour on the job is the decisive and most often neglected level.
- Measurement starts before the training: no baseline, no comparison.
- An ROI needs a measurable result and cautious attribution.
- Not everything has to be expressed in money — but everything should have a goal.
Why is impact so hard to measure?
Training works through people, and people work in an environment that keeps changing. If sales rise after a sales training, that may be due to the training — or to a price cut, a new product, a good season. This ambiguity tempts teams into two extremes: not measuring at all, or crediting every positive development to training. Both damage L&D's credibility.
An honest approach measures on several levels, sets goals in advance and is open about what a number does and does not say.
The four-level model
The most widely used framework is the four-level model, which goes back to work from the 1950s and is often named after its originator as the Kirkpatrick model. It distinguishes four levels that build on each other.
| Level | Question | Typical measurement | Effort |
|---|---|---|---|
| 1 Reaction | How was the training experienced? | Short survey right afterwards | Low |
| 2 Learning | What was learned? | Test before and after, exam, practical task | Medium |
| 3 Behaviour | Is it applied at work? | Observation, conversation with the manager, spot checks after weeks | High |
| 4 Results | What changed for the business? | Metrics such as error rate, lead time, accidents, resignations | High |
How do you measure behaviour on the job?
Level 3 is the most important and the one most often skipped. This is where it is decided whether training was more than a pleasant day. Simple means work well: an agreed conversation between employee and manager a few weeks after the training, a concrete application task as the last step of the learning path, spot checks of work results, or a short survey of colleagues.
It helps to record before the training which behaviour should change. “Can close complaints without asking for help” can be observed; “is more customer-focused” cannot.
Skill levels can help too: when a manager confirms that someone now masters a skill at the level “Applying”, that is an indication at level 3 — provided the level is described in observable behaviour.
Calculating ROI honestly
Return on investment relates the benefit of a measure to its cost: (benefit minus cost) divided by cost. It is often described as a fifth level on top of the four-level model. The formula is simple; the numbers behind it are not.
Costs must be captured completely: not just fees and licences, but participants' working time, the time of trainers and managers, travel and administration. On the benefit side you need a measurable quantity at level 4 and cautious attribution. Where possible, compare with a group that has not yet taken part. Where that is not possible, ask those involved to estimate what share of the change is due to the training, and discount that estimate further.
An honest ROI states its assumptions. A figure like “1:4” without derivation does more harm than good — at the latest when finance starts asking.
When is an ROI calculation worth it?
Not every measure needs a profitability calculation. For mandatory training the question is the wrong one: it is required, and its benefit is legal certainty and protection. What counts here is completion rate, overdue items and how verifiable the records are. For large, expensive programmes with a clear business goal — reducing scrap or faster onboarding — the effort pays off.
- Mandatory training: completion rate, overdue items, audit readiness.
- Onboarding: time to independence, early turnover, feedback from new hires.
- Professional development: skill levels before and after, error rates.
- Leadership development: employee survey, team turnover, quality of conversations.
What data a learning platform can — and cannot — provide
Learning platforms mainly deliver data for levels 1 and 2: completion rates, time spent, test results, feedback. That is useful but not yet proof of impact. Level 3 needs information from day-to-day work — a manager's confirmation, a completed practical task, a confirmed skill level. Level 4 needs business metrics, which usually live in other systems.
A platform can make the bridge easier if learning paths end with practical steps that someone confirms, and if skills are tracked with target and actual levels. In Universe's People Development area, for example, the HR overview shows running onboardings, mandatory training completion as a percentage and overdue assignments; onboarding feedback with one or two stars appears as an early warning. Connecting this to business metrics remains a task for analysis within the company.
Step by step
Set the goal
Before the training, define which behaviour and which metric should change.
Capture a baseline
Measure or document the situation before the training.
Measure levels 1 and 2
Right after the training, capture reaction and learning.
Check level 3
After a few weeks, observe application on the job or clarify it in a conversation.
Evaluate level 4
Compare metrics, name other influences, attribute the result cautiously.
Terms
- Four-level model
- A framework for evaluating training at the levels of reaction, learning, behaviour and results.
- ROI
- Return on investment: (benefit minus cost) divided by cost.
- Learning transfer
- Applying what was learned in actual work.
Frequently asked questions
Is a satisfaction survey enough to measure success?
No. It shows how training was experienced, not whether people learned and applied anything. It is useful as an early warning but does not replace measurement at the learning and behaviour levels.
What is a good ROI for training?
There is no universal target. What matters is that the calculation is traceable: full costs, measurable benefit, openly stated assumptions.
When should you measure?
Before, to have a baseline; right afterwards for reaction and learning; after a few weeks to months for behaviour and results.
How do you deal with influences outside the training?
With comparison groups where possible, otherwise with cautious estimates that are discounted further. The key is to disclose them.
How it works in Universe
Related articles
- Planning a training budget: method and cost typesHow companies plan their training budget: identify needs, capture all cost types, separate mandatory and development spend, keep a reserve and steer during the year.
- Building a learning culture: what companies can actually doA learning culture does not come from posters but from time, role models and a healthy attitude to mistakes. Concrete levers for leadership, HR and teams.
- Measuring onboarding success: metrics that mean somethingHow to tell good onboarding: early turnover, time to independence, feedback and plan completion — and how to interpret them honestly.
- How to build a skills matrix, with a level modelHow to build a skills matrix: choose skills, describe levels, set targets per role and turn the gaps you find into focused training — step by step.