All posts

Blog

Kirkpatrick Model: The 4 Levels of Training Evaluation Explained

November 6, 2026 · By Zohaib Butt

Kirkpatrick Model: The 4 Levels of Training Evaluation Explained

The training went fine. Attendance was good, the trainer got a round of applause, and now an executive asks the question every training owner dreads: did any of it work?

The Kirkpatrick model exists to answer that question without hand-waving. It is the four-level framework most training teams reach for first, and it matters more than ever now that US organizations spend $102.8 billion a year on training — about $874 per learner (Training magazine, 2023 Training Industry Report). Evaluation is how that spend stops being an act of faith.

This guide explains each of the 4 levels in plain English, shows what each one costs to run, maps the evidence you can collect, and tells you how far up the model your program should actually go.

Key Highlights

  • The Kirkpatrick model has 4 levels: reaction, learning, behavior, results (Kirkpatrick's framework; structural fact).
  • US organizations spent $102.8 billion on training in 2023, about $874 per learner (Training magazine, 2023).
  • Evaluation effort climbs with each level: about 2 hours for level 1 versus 35 for level 4 per program (illustrative planning midpoints).
  • Managers drive 70% of the variance in team engagement (Gallup, 2015), which is why level 3 evidence comes from them.
  • 94% of employees say they would stay longer at a company that invests in their learning (LinkedIn Learning, 2018).
  • In our planning model, practice plus refreshers hold skill checks near 72% at 90 days versus 40% without (illustrative).
  • Most programs should run levels 1–3 well before chasing a level-4 claim (author's position).

What the Kirkpatrick Model Is

The Kirkpatrick model is a four-level framework for evaluating training, first laid out by Donald Kirkpatrick in the 1950s and refined by the Kirkpatrick family since. Level 1 measures reaction, level 2 measures learning, level 3 measures behavior, and level 4 measures results. The order matters: each level gets harder to measure and more valuable to know.

The model's genius is the question it forces at every level. Reaction asks whether it was worth their time, learning asks whether they can now do it, behavior asks whether they do it on the job, and results asks whether the business moved. Those are four different questions with four different kinds of evidence, and programs that blur them end up with one survey pretending to answer everything.

Bar chart of evaluation effort per Kirkpatrick model level in hours per program: level 1 reaction 2, level 2 learning 6, level 3 behavior 20, level 4 results 35. Illustrative midpoints from the author's 2026 planning notes.

Level Question it answers Typical evidence When to run it
1. Reaction Was it worth their time? Short survey, live questions During and right after
2. Learning Can they do it now? Skill checks, scored quizzes Before and at the end
3. Behavior Do they do it on the job? Observation, manager reports 30–90 days out
4. Results Did the business move? Business metrics, paired with finance 1–2 quarters out

The effort chart is our planning reality (illustrative midpoints): we plan about 2 hours per program for level 1, 6 for level 2, 20 for level 3 and 35 for level 4. That cost curve is exactly why programs should choose their levels deliberately rather than by ambition.

Key takeaway: know what question each level answers before you decide which levels to run.

Level 1: Reaction — Worth Their Time?

Level 1 is the smile sheet, and it is useful when you keep it short and honest. Ask whether the session was relevant to their actual job, whether the pace worked, and what one thing they will apply this week. Keep the whole thing under five minutes; response quality collapses on long forms.

A five-point card takes about 2 minutes to complete, an eight-question form about 4, and an open-comment format pushes 8 (illustrative planning midpoints).

Horizontal bar chart of minutes learners will give a post-course survey: five-point card 2, eight-question form 4, open-comment form 8. Illustrative planning midpoints, author model 2026.

The trap at level 1 is treating comfort as learning. A session can be enjoyable, well-catered and completely ineffective, which is why reaction is only the first gate and never the verdict. Ask the "apply this week" question and read the answers; they are your cheapest early-warning system.

Key takeaway: keep reaction surveys under five minutes and mine the apply-this-week answers.

Level 2: Learning — Can They Do It Now?

Level 2 asks whether knowledge or skill actually changed, and it needs a measurement, not a feeling. Run the same short skill check before the course and at the end, then repeat it later to see what stuck. The before-and-after gap is your learning claim; the later drop is the forgetting curve arriving on schedule.

In our planning model for a skills course: without a practice plan the cohort starts near 48%, ends near 84%, and lands at roughly 40% at the 90-day spot check; with practice assignments and spaced refreshers it ends near 86% and holds 72% at day 90 (illustrative midpoints).

Grouped bar chart of illustrative Kirkpatrick model level 2 skill-check scores in percent: no practice plan scores 48 at baseline, 84 at the end of course and 40 at the 90-day spot check, practice plus refreshers 48, 86 and 72. Illustrative author model, 2026.

Evidence type What it shows Effort to collect
Pre/post skill check The learning gap you caused Low, reusable per cohort
Scenario quiz Whether judgment transferred Medium, needs writing
Manager spot question Whether the skill is visible Low, one question

The fade after a course is expected, not a scandal — our guide to the forgetting curve explains the mechanics and the fix. What level 2 gives you is the honest baseline that makes the fade visible and fixable.

Key takeaway: run the same check before, at the end and at 90 days; three points tell the whole story.

Level 3: Behavior — Do They Do It on the Job?

Level 3 is where training budgets are actually justified or wasted, and it is the level most programs skip. The question is blunt: 30 to 90 days later, is the behavior happening? Evidence comes from observation, manager reports and short multi-rater checks rather than from the training platform.

Managers are the lever here. Managers account for 70% of the variance in team engagement (Gallup, State of the American Manager, 2015), and the same managers are the ones who see — or never notice — the new behavior. Put two prompt questions into their one-on-ones and the observation problem mostly solves itself.

Line chart of illustrative adoption of a new leadership habit in percent: no manager follow-up falls from 70 at week 2 to 45 at week 6 and 30 at week 12, manager prompts with refreshers hold 70, 60 and 55. Illustrative author model, 2026.

In our planning model, about 70% of managers try a new leadership habit in the first two weeks after a program. Without follow-up it slides to roughly 45% by week 6 and 30% by week 12, while manager prompts and refreshers hold 60% and 55% (illustrative midpoints). For structured multi-rater evidence, our 360-degree feedback guide covers the mechanics.

Key takeaway: level 3 lives with the managers, so build the prompts into their routine, not into a report.

Level 4: Results — Did the Business Move?

Level 4 connects training to business metrics, and honesty is the whole game here. Pick one or two metrics the trained behavior could plausibly move — rework, ramp time, overtime, quality escapes, retention — and baseline them before the program. Then let the numbers speak, including when they say nothing.

A working example we use to size the prize (illustrative, composite of client patterns): a 40-person support team cuts average rework after a coaching-skills program. Rework avoided runs about $3,000 a month, faster ramp-up of new hires about $2,500, and avoided overtime about $1,800. Against a program cost of a few thousand dollars, that is a defensible results claim; without the baselines, it is a story.

Horizontal bar chart of illustrative monthly value from one team behavior change in USD: rework avoided 3,000, faster ramp-up 2,500, overtime avoided 1,800. Illustrative midpoints from the author's 2026 client models.

Two disciplines keep level 4 honest. Attribute conservatively — training shares credit with everything else that changed that quarter. And connect the results story to money with the discipline from our training ROI formula guide, so the claim survives a finance review.

Key takeaway: baseline the metric before the program; a results claim without a baseline is an anecdote.

How Far Up the Kirkpatrick Model Should You Go?

Run all four levels for your biggest programs and stop where the cost stops making sense for the small ones. Our rule of thumb with US mid-market teams: every program gets levels 1 and 2, programs touching more than 25 people or a leadership cohort get level 3, and level 4 is reserved for the two or three programs a year that carry a real budget line.

Your situation Stop at Why
Half-day refresher, 10 people Level 2 The skill check answers the question
Process rollout, 40 people Level 3 Behavior is the deliverable
Leadership cohort, 12 managers Level 3 Manager habits drive the team numbers
Flagship program with a budget line Level 4 Finance will ask, so answer first

One more honesty note: the model has critics, and the loudest complaint is that levels run in sequence when business results should drive the design backward. That criticism is fair and compatible with the framework — start from the result you need, then work down the levels to build the chain. The model is a checklist, not a religion.

Key takeaway: pick your evaluation depth by program size and budget, and design from level 4 backward.

What Your Business Should Do

Days 1–30: Baseline what you will evaluate.

  • Pick your two biggest programs for the next quarter and write each level's question.
  • Build one short skill check per program and run it as the pre-test.
  • Name the level-3 behavior each program owes you, in one sentence each.

Days 31–60: Collect level 1–3 evidence while programs run.

  • Keep reaction surveys under five minutes and read the apply-this-week answers.
  • Schedule the 90-day repeat of each skill check before the course starts.
  • Write the two manager-prompt questions that watch for the target behavior.

Days 61–90: Report up the ladder and choose your depth.

  • Repeat the skill checks and report the before, after and day-90 numbers.
  • Ask managers for one observable behavior change per participant.
  • Choose which flagship program earns a level-4 metric and baseline it with finance.

The Bottom Line

The Kirkpatrick model earns its keep because it separates four questions that most programs blur into one survey. US organizations spend $874 per learner a year (Training magazine, 2023), and evaluation effort runs from about 2 hours at level 1 to 35 at level 4 in our planning model (illustrative). Run levels 1–2 on everything, add level 3 where behavior is the deliverable, and reserve level 4 for the flagship programs finance actually funds. If you want this mapped to your own team, book a free strategy call or start with the free resource library.

Work With Zohaib Butt

1. Free Resource Library — Zohaibbutt.com/store

World-class training and coaching resources, all in one place. Download the assessments, templates and playbooks you need today.

2. Online Courses — learning.zohaibbutt.com

Self-paced leadership and coaching programs for managers and owners who want structure, not scattered tips.

3. Free Strategy Session — funnelsthatsell.com

Book your complimentary strategy call and we will map your biggest people-development bottleneck to a 90-day fix.

Free assessment pack (PDF)

Plan what each level will ask before the program starts

Download the PDF pack

FAQ

What are the 4 levels of the Kirkpatrick model?

They are reaction, learning, behavior and results. Level 1 asks whether the training felt worth the learner's time, level 2 asks whether they can now do the skill, level 3 asks whether they do it on the job 30–90 days later, and level 4 asks whether a business metric moved. Each level costs more effort and answers a more expensive question.

Is the Kirkpatrick model still relevant in 2026?

Yes — with one adjustment. The four questions remain the right questions for any program spending real money, which across US organizations meant $102.8 billion in 2023 (Training magazine). The modern adjustment is to design from level 4 backward: pick the business result first, then build the behavior, learning and reaction chain underneath it. Critics who say results should drive design are right, and the model can carry that.

What is an example of Kirkpatrick level 4 evaluation?

A leadership program for 12 managers whose team retention was baselined at the start. Six months later, voluntary exits on the coached teams run lower and promotion readiness scores run higher, and finance co-signs the numbers. The claim stays honest because the baseline came first and the metric belongs to the business, not to the training platform.

How do I apply the Kirkpatrick model with a small training budget?

Run levels 1 and 2 on everything, because they cost hours, not days: a five-minute survey and a reusable skill check. Add level 3 with manager prompts rather than formal observation. Skip level 4 until a program is big enough that finance would ask anyway — that is the point where results measurement starts paying for itself.