# Kirkpatrick Model: Four Levels and a Practical Evaluation Plan

Use the four Kirkpatrick levels to build a practical training evaluation plan, with evidence sources, timing, and limits on claims about learning and impact.

- Canonical URL: https://www.harmate.com/en/blog/kirkpatrick-model-four-levels-evaluation-plan
- Author: Harmate Team
- Published: 2026-10-25
- Updated: 2026-10-25T22:49:13.294452+00:00
- Language: en

## Content

# Kirkpatrick Model: Four Levels and a Practical Evaluation Plan

The Kirkpatrick model distinguishes four dimensions of training evaluation: participants’ reactions, learning, workplace behavior, and organizational results. It helps you specify what you want to examine. A positive course evaluation does not, by itself, establish that people learned, applied their learning, or improved business results.

Start with a decision: should you change a practice activity, provide support on the job, or continue a program? Then identify the evidence needed, who can provide it, and when it becomes available. The worked plan below follows one training program across the four levels without treating a survey as a complete evaluation system.

## What the four levels cover

[Kirkpatrick Partners presents the model](https://www.kirkpatrickpartners.com/) using the terms Reaction, Learning, Behavior, and Results. These are different evaluation dimensions. Their numbering does not demonstrate an automatic causal chain from one to the next.

**Level 1: Reaction.** How did participants experience the training and its relevance? Comments about a realistic practice task, unclear instructions, or insufficient practice time describe that experience. Enjoying an activity does not establish that someone can perform the task.

**Level 2: Learning.** What evidence can you collect about learning? The assessment should match the objective. Explaining a rule, solving a case, preparing a document, and performing a procedure require different tasks and criteria. A report of confidence provides different information from work assessed against explicit criteria.

**Level 3: Behavior.** What happens when people have an opportunity to apply their learning at work? Separate reported use, available work samples, and direct observation. A person may not have tried a technique because the relevant task never arose, a resource was unavailable, or workplace processes prevented it. None of those explanations establishes a failure to learn.

**Level 4: Results.** How are the intended organizational outcomes changing? Define the outcome with the sponsor and establish whether relevant data are available. Improvement after training can have other explanations, including a new tool, a process change, different workloads, or manager support. Attributing an effect to training requires an appropriate causal evaluation design.

The [University of Strasbourg’s continuing education service](https://sfc.unistra.fr/formation-continue-de-luniversite-de-strasbourg/notre-offre/kirkpatrick/) also distinguishes positive reactions from learning and highlights workplace conditions for transfer. Preserve those distinctions in both collection and reporting.

## Worked example: training staff to hand off customer cases

**This example is entirely fictional.** A training provider works with a team that passes customer requests to another department. The sponsor wants to reduce avoidable requests for clarification. The course focuses on documenting the original request, checks already completed, and the next action needed. No customer, measurement, or outcome below represents a real engagement.

The table is an original evidence-planning worksheet, not an official Kirkpatrick form.

| Dimension | Evaluation question | Proposed evidence and timing | Conclusion the evidence does not justify |
| --- | --- | --- | --- |
| Reaction | Did the practice case resemble real handoffs? Which difference mattered? | An open response at the end of the course; participants can describe a different situation without customer details | A well-liked case proves that participants can perform the task |
| Learning | Does a handoff for a new scenario state the request, completed checks, and next action? | A work sample assessed against shared criteria; initial and final tasks compared only when their demands and conditions support comparison | One successful sample demonstrates competence in every workplace situation |
| Behavior | During an actual handoff, what was used or adapted? What prevented use? | Follow-up after an actual opportunity; a de-identified sample if permitted, or an agreed observation | A self-report establishes regular use across the entire team |
| Results | Are avoidable clarification requests changing within the agreed scope? | Comparable operational records, an agreed definition and period, and documentation of other changes | A before-and-after change alone establishes training’s causal effect |

Assign responsibility before collecting anything. The trainer may prepare assessment tasks and gather feedback. The operational team needs to confirm access to work samples and define what counts as an avoidable clarification request. If a source will not be available, narrow the intended conclusion in advance.

Do not collect complete customer records simply to make the report look more convincing. Keep only the necessary information, remove identifiers, and agree who can access the material. A contextualized excerpt may support the discussion without exposing a whole case.

## Match the collection method to the question

For immediate reactions, try an open prompt such as: “Which part of the exercise helped you clarify a handoff, and which part did not fit your work?” It can surface situations the trainer did not anticipate. That richness helps explain experience; it does not directly assess learning or guarantee representative feedback. A closed question can support measurement or comparison when its wording and scope remain clear. Our [training satisfaction questionnaire guide](/en/blog/training-satisfaction-questionnaire-why-your-results-are-often-unusable) develops this part of the process.

For learning, ask participants to produce something aligned with the objective rather than only rate their confidence. In this example, the trainer reviews a handoff document, records which criteria are met, and notes the help provided. A final task describes performance under those conditions. A claim about progress needs an appropriate comparison, not two tasks with unknown differences in difficulty.

For workplace application, begin with the opportunity: “Since the session, have you prepared a handoff? Describe one you are comfortable discussing.” If no opportunity occurred, do not force a judgment about use. If it did, ask what was tried, adapted, and supported. The [post-training evaluation guide](/en/blog/post-training-evaluation-questions-and-method-for-assessing-workplace-application) provides a more detailed approach to that follow-up.

## What to do when the evidence disagrees

Suppose participants like the training, but several work samples omit the next action required. You might revise that part of the practice activity. A high satisfaction rating should not erase the observation.

Now suppose work samples during training meet the criteria, but later feedback describes software fields that do not accommodate the handoff format. That calls for a discussion of the workflow with the sponsor. Repeating the course without examining the constraint may leave the original problem unresolved. These are illustrative decision paths, not automatic diagnostic rules.

If operational figures improve while a new procedure is introduced, report both facts. You may decide to retain a useful program without claiming a proven causal impact. Preserve contrary cases and response counts too: the people who answered may have different experiences from those who did not.

## Keep the evidence boundaries in the report

Give each finding a source, period, scope, and limitation. Distinguish a participant’s account from an observed work sample and an operational measure. An undocumented dimension remains undocumented; it is neither a zero result nor a level you can mark as automatically achieved. Our [training evaluation report example](/en/blog/training-evaluation-report-a-complete-example-for-a-client) shows how to present those distinctions to a sponsor.

When you have many written responses, [Harmate’s open-question analysis](/en/product/open-questions) supports reviewing themes and findings alongside source excerpts, then selecting material for a decision file. Interpretation remains with the trainer and sponsor. Reviewing comments does not replace a learning assessment or an impact evaluation design.

An evaluation plan is ready when every intended conclusion has an accessible source and an explicit boundary. That lets you agree what the evaluation can establish before asking participants to contribute their time.