How to Evaluate Training: A Provider's Guide to Kirkpatrick

To evaluate training as a provider, agree with the client what result the course is meant to produce, measure reaction and learning yourself on every course date, agree who will check behaviour and results back at work, and report all of it to the client on one page. The Kirkpatrick model gives the structure: four levels, planned from the result back to the course. Most evaluation advice is written for an in-house learning team evaluating its own staff. A provider’s position is different, because the learners work for someone else and the evidence for the higher levels sits in the client’s business. A provider needs to know what it can measure on its own, what it needs the client for, and how to make evaluation part of the sale.

Key takeaways

• Kirkpatrick’s four levels are reaction, learning, behaviour and results. A provider can measure the first two alone; the last two need the client.
• Plan from Level 4 backwards: agree the result and the behaviours that lead to it before the course is designed or sold.
• A learner’s own report of using the course is still a Level 1 measure. Behaviour needs another source, such as the learner’s manager.
• Ask about relevance, confidence and intent to apply, with answer choices that mean something, rather than a row of agree/disagree scales.
• Report back to the client on one page after every course, and again after the follow-up. It is the evidence behind the next order.

Why a provider evaluates at all

Three groups have a stake in whether a course worked: the client who paid for it, the regulator or standards body where the training is regulated, and your own team. Each wants different evidence, and each does part of the work. The table splits what falls to you from what you need the others to do.

Who	What they want	What you do	What they do
The client’s L&D or HR lead, and the learners’ managers	Evidence the spend worked, in a form the buyer can show their own manager	Agree the result at the sale; run the feedback form and assessments on every date; send a one-page summary after each date and after the follow-up	Name the result they want; have managers check the agreed behaviours back at work; share the business data they already track
A regulator or standards body, where the training is regulated	A documented evaluation procedure, and evidence that you act on what it finds	Write the procedure down; collect feedback from learners, and from employers where the rules ask for it; record what you changed	Set the requirements, and check them at audit or accreditation, or when an employer chooses a provider
Your own team	Which trainer, exercise and course date worked, and which needs changing	Read the answers within days; share each trainer’s feedback with them; compare dates and trainers over time	Trainers adjust their sessions and tell you what they changed

Most of the method falls to you, because many clients have none of their own. In the CIPD’s 2023 Learning at Work survey of 1,108 UK organisations, only 50% agreed that they had a process for assessing learning impact. A provider who brings the method, and the first half of the evidence, takes that work off the buyer’s desk and gives them a reason to choose the same provider when the contract comes up again.

Where the training is regulated, these are the rules that ask for evaluation.

Where	What the rules expect
UK, first aid at work	HSE’s guidance for employers choosing a first-aid training provider says the provider should have “a documented course evaluation procedure that includes feedback from students”, alongside a documented quality assurance plan
Australia, registered training organisations	The 2025 Outcome Standards require that students “are supported to provide feedback” and that feedback is “used by the organisation to inform continuous improvement” (Standard 2.7), and that the organisation monitors and evaluates itself using data from students and employers (Standard 4.4)
US, IACET-accredited providers	The ANSI/IACET 1-2018 Standard for Continuing Education and Training has categories for learning outcomes, the assessment of learning outcomes and the evaluation of learning events
International	ISO 29993, the standard for learning services outside formal education, covers services where “the goals of learning are defined and the services are evaluated”

The Kirkpatrick model, from the provider’s side

Donald Kirkpatrick set out the four levels in a series of four articles that began in the November 1959 issue of the Journal of the American Society of Training Directors. The model has been refined since, most recently by Kirkpatrick Partners as the “New World” model, and it is still the frame most corporate learning teams use, which makes it the shared language for a provider talking to a client.

The table uses Kirkpatrick Partners’ current definitions, shortened, and adds a column for who holds the evidence, which is the question that matters most to a provider.

Level	What it measures	Who holds the evidence	When	How a provider gets it
1. Reaction	Whether learners find the course “favorable, engaging, relevant” and supportive of the outcome	You	At the end of the course	A feedback form sent when the learner completes
2. Learning	Whether learners acquire the intended “knowledge, skills, attitude, confidence, and commitment”	You	During and at the end of the course	Assessments, quizzes, practical checks, before-and-after questions
3. Behaviour	Whether learners perform the critical behaviours back at work, with support and accountability	The client	Once the skill should be in use	The client’s managers, observation, the client’s own records
4. Results	Whether the targeted organisational outcomes occur	The client	Over months	The client’s business data, shared with you

Two refinements in the New World model are useful to a provider. At Level 1, Kirkpatrick Partners call relevance the most important thing to measure: a course that learners do not see as relevant to their work is unlikely to be applied. At Level 2 they add confidence and commitment, which they describe as “early predictors” of what happens back at work. Both can be asked on the day, which makes them the closest thing a provider has to a Level 3 signal before the course is over.

Kirkpatrick Partners also say that “it is critical to start with Level 4”. The levels are numbered in the order they happen, and planned in the opposite order: the result first, then the behaviours that produce it, then what learners need to know, then how the course should feel.

Phillips’ fifth level: return on investment

Jack and Patti Phillips added a fifth level, return on investment, in the ROI Methodology. It converts the Level 4 results into money and compares them with the full cost of the programme:

ROI (%) = (programme benefits − programme costs) ÷ programme costs × 100

The step that makes or breaks the figure comes before the sum. The methodology requires isolating the effect of the training from everything else that changed at the same time, by methods that range from control groups and trend lines to estimates from participants and their managers. Without that step, a rise in sales after a sales course says nothing about the course.

What the model leaves out

The four levels have critics. Elwood Holton argued in 1996 that the levels are a taxonomy rather than a model, because nothing shows that one level causes the next. Will Thalheimer, who designed the Learning-Transfer Evaluation Model (LTEM) as an alternative, argues that the four levels put all learning into one bucket, where LTEM separates knowledge, decision-making and task performance. Kirkpatrick’s own reply, quoted in a 2004 TD magazine debate, was that people find the four levels helpful whatever you call them.

For a provider the practical point is narrower: use the levels as the shared language with the client, and take the criticism as a warning about what each level can and cannot prove.

What you can measure on your own, and what you can’t

Organisations measure the lower levels far more often than the higher ones. In ATD’s 2019 research with 779 talent development professionals, about eight in ten organisations evaluated at Levels 1 and 2, and the share fell at every level above.

The pattern has a simple cause. Levels 1 and 2 happen in your room, or your online course, where you control the questions and the timing. Levels 3 and 4 happen in the client’s workplace, weeks or months later, and depend on things you do not control: whether the learner’s manager expects the new behaviour, whether the tools and time to use it exist, and what else changed in the business.

The tempting shortcut is to ask learners afterwards whether they are using what they learned. That answer has value, but it is a learner’s opinion. As Thalheimer puts it about asking learners for Level 3 and 4 data: “we are getting their Level 1 reactions. We are NOT getting Level 3 data.” Treat a learner’s report of using the course as a lead to check, and get the evidence for behaviour from someone who can see it.

So the division of work is clear. You own Levels 1 and 2 on every course. For Levels 3 and 4 you own the method, the timing and the report, and the client supplies the evidence. That division is worth agreeing in writing before the first course date.

Agree the measures when you sell the course

The time to plan evaluation is the sale, because that is when the client explains why they are buying, and the result they want is the starting point for Level 4. A short evaluation agreement in the proposal turns that conversation into something both sides can check later.

What to agree	The question to ask	Example (illustrative: a manual handling course for a warehouse client)
The result	What should be different in the business if this works?	Fewer handling injuries in the client’s incident log
The critical behaviours	What will people do differently at work?	Team lifts used for heavy loads; loads checked before lifting
Who observes them	Who sees the work every day?	Shift supervisors
How and when	What will they check, and by when should it show?	A short observation checklist, completed once the course has had time to bed in
What the client shares	Which of their data will they give you, and at what level of detail?	Incident counts for the teams trained, before and after
What you report	What will you send, and when?	A course summary after each date, and a follow-up summary after the observation
Support back at work	What will the client do to make the behaviour stick?	Supervisors brief the team on the checklist before it starts

The last row is easy to skip, and it matters most. Kirkpatrick Partners define Level 3 as performing the behaviour while being “supported in and accountable for” it. If nobody back at work expects the new behaviour, the course gets the blame for an outcome the workplace decided. Agreeing the support in advance protects the client’s investment and your reputation.

Keep the agreement proportionate. For an open course with learners from many employers, Levels 1 and 2 are usually enough. For a private contract, the agreement can be half a page. For a large programme with a budget to defend, it is worth agreeing whether the client wants a return-on-investment figure before anyone promises one.

A feedback form that predicts something

The end-of-course form is where most evaluation starts and stops, and the usual version tells you very little. Thalheimer’s 2026 GROWLE report cites two meta-analyses that found traditional “smile sheets” correlate with learning at r = .09, which he calls “essentially zero”. The same report found that only 19% of learning professionals surveyed called their learner survey data “very useful”.

The fix is in the questions and the answer choices.

Start with what predicts use. Kirkpatrick Partners single out relevance, confidence and commitment, so ask whether the course fits the learner’s job, how confident they are to use it, and what they intend to do with it.

Write the questions about the learner rather than the trainer. Ken Phillips’ guidance on Level 1 forms, published by Kirkpatrick Partners, recommends learner-centred items, at least one question on how relevant the course was to the learner’s job, and only questions whose answers you would act on.

Then give the answers some meaning. An agree/disagree scale invites everyone to tick “agree”. Thalheimer recommends “distinctive questions”, where each answer describes a different state and the learner has to place themselves. Ken Phillips still recommends rating scales; if you use them, keep one scale with an odd number of points throughout, with words at each end. Finish with one open question about what would have made the course more useful, which gives you specific things to change.

Here is a short question bank in that spirit. Use five or six, not all of them.

Asks about	Question	Answer choices
Relevance	How much of today’s course applies to the work you do?	Almost none of it · Some of it, with changes · Most of it · Nearly all of it, as taught
Understanding	How well do you understand what was covered?	I’m still confused about key points · I understand the basics but not how to apply them · I could apply it with some help · I could apply it now without help
Confidence	How confident are you to use this at work next week?	Unsure I could use it · Confident with support · Confident on my own · Confident enough to show a colleague
Commitment	What will you do with this in the next month?	Nothing yet · Try one thing · Use it in my regular work · Change how my team does it
Barriers	What could stop you using it?	Open text
The course	What one change would have made this course more useful to you?	Open text
The trainer	How would you rate the trainer?	1 to 5 stars, with a reason

Send the form when the learner completes, while the course is fresh, and read the answers within days. The CPD Standards Office’s guidance for trainers makes the same point. Reading evaluation sheets soon after a session gives trainers timely feedback, and participants should hear about any changes made in response.

Measuring learning

Level 2 is where a provider has the most control and the most to show a client. Use the measure that matches what the course teaches.

• For knowledge, ask the same short set of questions before and after the course, so the report shows the change as well as the end score. Quiz pass marks and retake counts belong in the course record.
• For decisions, use scenario questions, where the learner chooses what to do in a realistic situation. They test whether knowledge turns into judgement, which is why Thalheimer’s LTEM separates decision-making competence from simple recall.
• For skills, a practical check against a short list of observable steps, completed by the trainer, gives the clearest Level 2 evidence on a hands-on course.
• For confidence and commitment, the two Level 2 additions in the New World model, ask on the feedback form.

For an accredited course, the assessment set by the awarding or accrediting body already does part of this job. Report its results alongside your own measures, so the client sees the full picture for their people. Training certificates and CPD covers what to record when a learner passes.

The follow-up after the course

Level 3 needs a date, a method and an owner, all agreed at the sale.

The right date for the follow-up depends on how soon the skill should show. A course on a system the learners use daily can be checked within weeks; a course on annual appraisals has to wait for the appraisal round. Agree it with the client rather than picking a standard interval.

Make the follow-up easy to complete. Ask managers for a handful of observable items per person rather than a written report. A delayed survey to the learners, which Thalheimer recommends alongside the end-of-course form, adds what got in the way, and is useful as long as it is read as the learners’ view.

Results and return on investment

Level 4 is the client’s data, and the provider’s job is to ask for it at the start and report what it shows at the end. Choose a measure the client already tracks, such as errors, complaints, audit findings, time to competence for new starters, or sales, so nobody has to build a new report.

The ROI Methodology’s guiding principles are worth following even when you never calculate a return. Isolate the effect of the training before you claim it, and when an estimate is needed, take the most conservative one. A provider who says “the incident rate for the teams trained fell, and the client’s managers estimate the course accounted for some of that” is more credible than one who claims all of it.

A full return-on-investment study costs time on both sides, and most organisations do not run one: in the ATD research, 16% evaluated at Level 5. Offer it for large programmes where the client needs a financial case, and agree the method before the programme starts.

Report back to the client

The report is where evaluation turns into the next order. Send a short summary after every course date, and a second one after the follow-up. Keep it to one page, so the person who bought the course can forward it without editing it.

Section	What goes in	Where it comes from
Who took part	Learners booked, attended and completed, by team if the client wants that	Your registrations and attendance
Reaction	Relevance, confidence and commitment results, and three or four representative comments	The feedback form
Learning	Assessment and quiz results, before-and-after change, practical checks passed	Your course records
Behaviour	What the managers observed at the follow-up, and what the learners said got in the way	The follow-up
Results	The agreed measure, before and after, with the caveat about other causes	The client’s data
What happens next	What you will change in the course, and what the client could do to support the behaviour	Your judgement

Write the summary for someone who was not in the room. Lead with the result, keep the comments verbatim, and put the actions at the end. How to sell training to corporate L&D buyers covers the rest of the client relationship that this report supports.

Use the data on your own courses

The same answers that go into a client report tell you where to improve. Compare course dates, trainers and formats over time rather than reading each form in isolation, and look for the questions where the answers shift. If relevance drops on one course date, check who was booked and whether the examples fitted their work. If confidence drops, check whether the practice time was cut.

Record what you changed and why. For a regulated provider, that record is the continuous improvement the standards ask for. For everyone, it is what you show the next client who asks how you know your courses work.

Where Cademy fits

Evaluation works when the questions are asked at the right moment, every time, and the answers sit on the same record as the attendance, the assessment and the client. In Cademy, the feedback form is part of the course, and the request goes out when the learner completes it.

Asked at the right moment

You build a feedback form for a course with your own questions: open text, preset answer choices, a five-star rating, or a numbered scale with labels at each end. One form can serve many courses, or be copied and adapted for one. The request goes automatically when a learner completes: on a scheduled course once their sessions are marked attended, on an on-demand course once the lessons are done, and on a blended course when both are. If attendance is marked a few days late, the request still goes.

For the follow-up, you send the form to one learner or several, resend it to anyone who has not answered, and change the wording of the email. Each learner gets a personal link that works without logging in.

Read against the record

Responses appear in a table with one column per question, filtered by course, course date or trainer, and export to a spreadsheet with ratings and scores as numbers. Each form has an analytics tab with views, responses and the share of views that turned into a response, so you can see your response rate by period. Trainer ratings sit alongside the course rating, and System Admins see them for the dates each trainer taught.

admin.cademy.io/courses/emergency-first-aid-at-work/home

CoursesEmergency First Aid at WorkSend InvitesShareEdit Course

Emergency First Aid at Work

Leeds£245EFAW

Public

Upcoming DatesPast Dates

Sep30

Wed, Sep 30, 2026, 9:00 AM

Leeds, Room 2 9 / 12

Oct8

Thu, Oct 8, 2026, 9:00 AM

Doncaster 0 / 12

Oct14

Wed, Oct 14, 2026, 9:30 AM

Online 5 / 16

Create Date

OverviewOrders 7Registrants 9Reviews 3Tasks

OrdersRegistrantsCreate

Name	Status	Attendance
ARAnna Reid	Confirmed	Attended
TRTom Reid	Confirmed	Attended
LALena Adeyemi	Confirmed	Attended
RWRuth Walsh	Confirmed	Absent
DPDan Price	Confirmed	Attended

ResponsesParticipantsSep 30Export

Name	Applies to your work	Confidence to use it	Trainer
ARAnna Reid	Nearly all of it	Confident on my own	★★★★★
TRTom Reid	Most of it	Confident with support	★★★★★
DPDan Price	Most of it	Confident enough to show a colleague	★★★★★

Exported to a spreadsheet · ratings and scores as numbers

ResponsesParticipantsSep 30Resend Request

Name	Request	Response
ARAnna Reid	Sent on completion, Sep 30	Responded
TRTom Reid	Sent on completion, Sep 30	Responded
LALena Adeyemi	Sent on completion, Sep 30	Waiting
DPDan Price	Sent on completion, Sep 30	Responded

Tasks

Send joining detailsMon 28

Confirm Room 2 with facilitiesFri 25

Notes

Room 2 is booked from 08:30. Kit list in the cupboard by the door.

Level 2 sits on the same record. Quizzes carry a pass mark and retakes, and each registration shows its completion status from the attendance and the online lessons. A private date for a client is linked to that client’s organisation, so the people, the results and the feedback for a contract can be read together when you write the summary.

If you collect public reviews, the same form can include a review section, published automatically or after you approve each one, or you can leave it out and keep the feedback for your team.

To see your own feedback form set up this way, book a call and bring the form you use now, or start for free and send it after your first course date. Training management system shows the rest of the record behind each course date.

Frequently asked questions

What are the four levels of the Kirkpatrick model? Reaction, learning, behaviour and results. Reaction is whether learners find the training favourable, engaging and relevant. Learning is whether they gain the intended knowledge, skills, attitude, confidence and commitment. Behaviour is whether they apply it at work. Results is whether the targeted outcomes for the organisation occur.

Who created the Kirkpatrick model? Donald Kirkpatrick, in a series of four articles that began in the November 1959 issue of the Journal of the American Society of Training Directors. Kirkpatrick Partners now maintain the model, in a form they call the New World Kirkpatrick Model.

What is the difference between the Kirkpatrick and Phillips models? The Phillips ROI Methodology keeps Kirkpatrick’s four levels and adds a fifth, return on investment. It converts results into money, compares them with the full cost of the programme, and requires the effect of the training to be isolated from other causes first.

How do you calculate the ROI of training? Subtract the full programme costs from the programme benefits, divide by the programme costs and multiply by 100. The ROI Methodology requires the benefits to be isolated to the training and estimated conservatively before the sum is done.

Can a training provider measure Kirkpatrick Levels 3 and 4? Only with the client. Behaviour and results happen in the client’s workplace, so the evidence comes from the client’s managers and data. The provider can design the method, set the timing and write the report, and should agree all three at the sale.

What questions should a training feedback form ask? How relevant the course is to the learner’s work, how well they understand it, how confident they are to use it, what they intend to do with it, what could stop them, and what one change would have made it more useful. Use answer choices that describe distinct states rather than agree/disagree scales.

When should you send a training feedback form? Send it as the learner completes the course, while it is fresh, and read the answers within days. Add a short delayed survey once the skill should be in use, alongside the managers’ observation, to learn what learners have applied and what got in the way.

Do first aid training providers need a course evaluation procedure? In the UK, HSE’s guidance for employers selecting a first-aid training provider says the provider should have a documented course evaluation procedure that includes feedback from students, and a documented quality assurance plan. Employers are told to check that the provider can show it meets these criteria.

Do registered training organisations have to collect feedback? Yes. Under Australia’s 2025 Outcome Standards for registered training organisations, students must be supported to provide feedback, the feedback must inform continuous improvement, and the organisation must collect and analyse data that includes feedback from students and employers.

Is the Kirkpatrick model still used? Yes. In ATD’s 2019 research, about eight in ten organisations evaluated learning at Levels 1 and 2, 54% at Level 3 and 38% at Level 4. Critics, including Elwood Holton and Will Thalheimer, argue that the levels do not show cause and effect, and Thalheimer’s LTEM is one alternative.

Sources

Figures and quotations checked 16 October 2026 on the publishers’ own pages.

• Kirkpatrick Partners, The Kirkpatrick Model
• Kirkpatrick, D. L., Techniques for evaluating training programs, Journal of the American Society of Training Directors, series beginning November 1959 (ProQuest record)
• Kirkpatrick Partners, Eight tips on developing valid Level 1 evaluation forms (Ken Phillips)
• ROI Institute, Introduction to the ROI Methodology (J. J. Phillips and P. P. Phillips)
• ROI Institute, ROI Methodology process model and guiding principles
• ATD, L&Ds struggle with learning evaluation (Effective Evaluation research, 2019)
• CIPD, Learning at Work 2023 survey report
• Will Thalheimer, GROWLE Report 2026
• Will Thalheimer, One of the biggest lies in learning evaluation: asking learners about Level 3 and 4 (2018)
• Will Thalheimer, LTEM: the Learning-Transfer Evaluation Model
• Will Thalheimer, Performance-focused smile sheets (handout)
• Holton, E. F. (1996), The flawed four-level evaluation model, Human Resource Development Quarterly 7(1)
• ATD, TD magazine, The evaluation heavyweight match (2004)
• Health and Safety Executive, GEIS3: Selecting a first-aid training provider
• Outcome Standards for NVR Registered Training Organisations 2025 (Australia)
• IACET, ANSI/IACET 1-2018 Standard for Continuing Education and Training
• ISO 29993:2017, Learning services outside formal education
• The CPD Standards Office, Training cycle: evaluate
• Cademy: Forms: feedback and course reviews and training management system