On this page
- Why ROI is rarely measured properly
- The Kirkpatrick Model
- The Phillips ROI Methodology
- Does leadership training actually work?
- Why skills often don't transfer to the job
- What to agree before training starts
- Practical measures beyond a feedback form
- Calculating a basic ROI figure
- Common measurement mistakes
- Questions to ask providers
- FAQs
- Sources and further reading
Ask most organisations how they know their training worked, and the honest answer is a same-day feedback form asking whether people enjoyed the session. That measures satisfaction, not impact. Two established frameworks, Kirkpatrick and Phillips, set out what measuring actual impact looks like, and why so few organisations do it.
Why ROI is rarely measured properly
Isolating training's specific contribution to a business outcome from everything else happening at the same time, management changes, market conditions, seasonal variation, is genuinely hard. It also has to be planned before the training runs, not bolted on afterwards. A same-day feedback form is easy to collect; tracking a specific job-performance metric three months later requires deciding what to track in advance and actually following up. That gap, not a lack of available methodology, is the real reason most measurement stops at satisfaction scores.
The Kirkpatrick Model
Introduced by Donald Kirkpatrick in 1959, this remains the most widely used framework for evaluating training, structured across four levels of increasing rigour and decreasing ease of measurement.
The four levels of the Kirkpatrick Model
| Level | What it asks | How it's typically measured |
|---|---|---|
| Level 1: Reaction | Did participants find it engaging and relevant? | Same-day feedback form or survey |
| Level 2: Learning | Did knowledge or skill actually increase? | Pre/post assessment, skills test, demonstration |
| Level 3: Behaviour | Is it being applied back in the job? | Manager observation, follow-up survey, 90-day check-in |
| Level 4: Results | Did it move a real business outcome? | Performance data, error rates, retention, sales figures |
Level
Level 1: Reaction
What it asks
Did participants find it engaging and relevant?
How it's typically measured
Same-day feedback form or survey
Level
Level 2: Learning
What it asks
Did knowledge or skill actually increase?
How it's typically measured
Pre/post assessment, skills test, demonstration
Level
Level 3: Behaviour
What it asks
Is it being applied back in the job?
How it's typically measured
Manager observation, follow-up survey, 90-day check-in
Level
Level 4: Results
What it asks
Did it move a real business outcome?
How it's typically measured
Performance data, error rates, retention, sales figures
Note
The Phillips ROI Methodology
Jack Phillips extended Kirkpatrick's framework with a fifth level: Return on Investment, comparing the monetary value of a training programme's outcomes against its full cost, including design, delivery and participant time. You may see it claimed that “90-100% of training is evaluated at Level 1, but only 5-10% reaches Level 5” presented as a measured industry statistic. That framing isn't quite accurate: those figures are Phillips' own recommended allocation of evaluation effort (a guideline for where to focus resource-intensive ROI studies), not an audited measurement of what organisations actually do. A separate, genuinely empirical survey - ASTD/i4cp's “Value of Evaluation” study (2009, 704 organisations) - found real-world Level 5 usage closer to 16-18%, still a small minority, but meaningfully higher than the commonly repeated “5-10%” figure. Both points support the same underlying reality - most organisations never reach a full ROI calculation - but it's worth citing the right one for the right claim.
Does leadership training actually work?
Setting evaluation models aside, the strongest evidence on whether leadership and management training works at all comes from two peer-reviewed meta-analyses, not vendor case studies. Lacerenza et al. (2017, Journal of Applied Psychology), pooling 335 independent samples, found leadership training produces meaningful positive effects across reactions, learning, on-the-job behaviour and business results alike - concluding leadership training is “more effective than previously thought.” Their analysis also identified what actually improves results: a needs analysis beforehand, built-in feedback, multiple delivery methods with real practice opportunities, spaced sessions rather than a single one-off event, and face-to-face or on-site delivery over purely self-administered content.
An earlier meta-analysis, Collins & Holton (2004, Human Resource Development Quarterly), pooling 83 studies, reached a similar conclusion but added an important caveat: effect sizes were noticeably larger for knowledge gain than for on-the-job or organisational outcomes, and studies with weaker designs (no control group) tended to report bigger effects than more rigorous ones - a reminder to be sceptical of any single training case study reporting a dramatic result.
Buyer warning
Why skills often don't transfer to the job
A frequently repeated claim - that “only 10% of what's learned in training transfers to the job” - is not a real research finding. It traces to a rhetorical, uncited remark in a 1982 trade article, later mistaken for a study result and repeated ever since. The genuine research on training transfer is more useful and less catchy: Baldwin & Ford's (1988, Personnel Psychology) foundational review identified three factors that determine whether skills stick - the trainee's own characteristics, how the training itself is designed, and the work environment they return to. Burke & Hutchins (2007, Human Resource Development Review) built on this, naming specific work-environment factors that most affect transfer: whether someone actually gets the opportunity to use the new skill, manager support, peer support, and whether the organisation holds people accountable for applying it.
The practical implication: a provider can design a good session, but if there's no opportunity to practise the skill afterwards and no manager reinforcement, the training is unlikely to stick regardless of how well it was delivered on the day.
What to agree before training starts
Checklist: Before the training runs
Practical measures beyond a feedback form
- Pre- and post-training knowledge or skills assessment (Level 2)
- Manager observation or a structured follow-up conversation 6–12 weeks later (Level 3)
- 360-degree feedback comparing before and after, for behaviour-focused programmes
- A specific business metric already being tracked: error rates, sales figures, retention, time-to-competence (Level 4)
- A simple before/after comparison against a control group or a prior period, where possible
Calculating a basic ROI figure
At its simplest: (monetary value of the improvement, minus the full cost of the training) divided by the full cost of the training, expressed as a percentage. The hard part is rarely the arithmetic, it's honestly isolating how much of the improvement is actually attributable to the training rather than other factors. Where that isolation genuinely can't be done with confidence, it's more honest to report the behaviour change qualitatively than to force a precise ROI percentage that overstates certainty.
Common measurement mistakes
- Only collecting same-day feedback and treating it as proof the training worked
- Not deciding what to measure until after the training has already run
- Attributing every change in a metric to training, ignoring other likely causes
- Measuring immediately after training instead of allowing time for behaviour to actually change on the job
- Forcing a Level 5 ROI figure onto training where the outcome genuinely isn't measurable in monetary terms
Questions to ask providers
- What evaluation do you build into your programmes, and at which Kirkpatrick level?
- Can you share results, at Level 3 or 4, from a genuinely comparable past client?
- How would you help us track behaviour change after the session ends, not just on the day?
- What would a Level 5 ROI evaluation look like for this specific programme, and is it realistic for our scenario?
FAQs
What is the Kirkpatrick Model?
A four-level framework for evaluating training, introduced by Donald Kirkpatrick in 1959 and still the most widely used training evaluation model: Reaction, Learning, Behaviour and Results. Most organisations only measure Level 1, which tells you whether people enjoyed the session, not whether it worked.
What is the Phillips ROI Methodology?
An extension of the Kirkpatrick Model, developed by Jack Phillips, adding a fifth level: Return on Investment, comparing the monetary benefit of training against its full cost. Phillips' own guidance recommends organisations reserve a full Level 5 ROI calculation for only a small share of programmes, since it takes real, deliberate effort to isolate training's specific financial impact from everything else going on in a business - a genuine 2009 survey of 704 organisations found real-world Level 5 usage around 16-18% in practice.
Does leadership or management training actually work?
Peer-reviewed evidence says yes, on average. Two meta-analyses pooling hundreds of independent studies - Lacerenza et al. (2017) and Collins & Holton (2004) - both found leadership training produces meaningful positive effects, though results-level outcomes are smaller and less consistently measured than reaction or learning-level ones. Provider testimonials and case studies don't count as this kind of evidence, since they're self-selected and rarely have a comparison group.
Why don't more organisations measure training ROI properly?
Isolating training's specific contribution from other factors, market conditions, management changes, seasonal variation, is genuinely difficult, and doing it properly takes planning that has to start before the training runs, not after. It's easier to collect a same-day feedback form than to track a specific job-performance metric three months later, so most measurement stops at Level 1 by default rather than by decision.
Can training ROI always be expressed as a number?
Not reliably for every type of training. Skills with a clear output, sales figures, error rates, production speed, are more measurable than softer outcomes like leadership judgement or team cohesion. For those, a mix of qualitative evidence (manager observation, 360 feedback) alongside any measurable proxy is usually more honest than forcing a single ROI percentage.
What should I measure if I can't do a full ROI calculation?
At minimum, agree a specific behaviour or metric you expect to change before the training starts, then check it again weeks or months later, not just on the day. That alone puts you ahead of measurement that stops at a same-day satisfaction score.
Sources and further reading
- The Kirkpatrick Model, Kirkpatrick Partners
- A critical analysis of evaluation practice: the Kirkpatrick model and the principle of beneficence, Bates, R. (2004), Evaluation and Program Planning, 27(3)
- ROI Basics, ROI Institute
- Value of Evaluation: Making Training Evaluations More Effective, ASTD/i4cp (2009)
- Leadership training design, delivery, and implementation: A meta-analysis, Lacerenza, C.N. et al. (2017), Journal of Applied Psychology, 102(12)
- The effectiveness of managerial leadership development programs: A meta-analysis, Collins, D.B. & Holton, E.F. III (2004), Human Resource Development Quarterly, 15(2)
- Transfer of training: A review and directions for future research, Baldwin, T.T. & Ford, J.K. (1988), Personnel Psychology, 41(1)
- Training transfer: An integrative literature review, Burke, L.A. & Hutchins, H.M. (2007), Human Resource Development Review, 6(3)
Choosing a provider who builds evaluation in from the start makes this whole process easier. CompareHQ's guide on how to choose a corporate training provider covers what to ask about measurement before you sign.