What Is the Kirkpatrick Model? – The Four Levels Explained

Your course launched, completions are rolling in and the survey scores look great. Then someone from finance asks a simple question: did it work? Not "did people like it," but did anything change? You know there's a standard way to answer that, and you've probably heard the name, but you're not sure how to put it into practice. What is the Kirkpatrick model, and how do you actually use its four levels?

The Kirkpatrick model is a four-level framework for evaluating training: Level 1 measures learners' reaction, Level 2 measures what they learned, Level 3 measures whether their on-the-job behavior changed, and Level 4 measures the business results that followed. Donald Kirkpatrick introduced the levels in a series of articles in 1959, and they remain one of the most widely used approaches to training evaluation. The levels build on each other, but you don't measure them in order of importance. The most useful way to apply the model is to start planning at Level 4, decide which results matter, then work backward to the behaviors, learning and reactions that lead there.

What Are the Four Levels of the Kirkpatrick Model?

The four levels of the Kirkpatrick model are Reaction, Learning, Behavior and Results. Each one answers a different question, uses different data and happens at a different time after training.

You'll find the current version of the model, maintained by Kirkpatrick's successors, on the Kirkpatrick Partners website. It keeps the same four levels with some added guidance, covered below.

Level 1: Reaction

Level 1 measures how learners felt about the training: whether it was relevant, engaging and useful. It's the easiest level to measure, which is why nearly every organization does it. A good Level 1 survey asks about relevance to the job and intent to apply, not just whether the room was comfortable.

Level 2: Learning

Level 2 measures whether learners gained the intended knowledge, skills, attitudes, confidence and commitment. Quizzes count, but skill demonstrations count more. If the course teaches a procedure, the strongest Level 2 evidence is watching someone perform it.

Level 3: Behavior

Level 3 measures whether people apply what they learned back on the job. This is where most training succeeds or fails, and it depends heavily on the work environment. A well-designed course can't overcome a manager who says "forget what they told you in training."

Level 4: Results

Level 4 measures the outcomes the organization cares about, such as fewer safety incidents, higher customer satisfaction, faster time to proficiency or lower error rates. These are the numbers leadership already watches.

How Long Does Each Level Take to Measure?

Each level takes longer to measure than the one before it, because behavior and results take time to show up. Levels 1 and 2 can be measured the day training ends, while Level 3 usually needs weeks and Level 4 often needs months.

The chart below shows an illustrative example for a customer service program. The timings are assumptions for a typical program, not research findings, but the pattern holds for most training.

This lag explains a common pattern. Teams often stop at Level 2 not because they don't care about results, but because the project has ended by the time results arrive. Planning the later measurements before launch fixes that.

How Do You Use the Kirkpatrick Model on a Real Program?

You use the Kirkpatrick model on a real program by planning from Level 4 down, then measuring from Level 1 up. Here's what that looks like for an example program teaching call center agents to resolve billing complaints on the first call.

1. Level 4 target. Raise the first-call resolution rate for billing complaints, which the operations team already tracks.
2. Level 3 behaviors. Agents verify the account, explain the charge in plain language and apply the right adjustment code without escalating.
3. Level 2 learning. Agents can choose the right adjustment code in ten realistic scenarios and explain a sample bill correctly.
4. Level 1 reaction. Agents rate the training as relevant to the calls they actually take and say they feel ready to use it.

Now the design follows naturally: practice scenarios mirror real calls, and supervisors get a short checklist to spot the target behaviors in call reviews. The baseline resolution rate is recorded before launch, and each level has a measure, an owner and a date.

If you can't name the Level 4 result before the course is built, the course probably doesn't have a clear purpose yet.

What Is the New World Kirkpatrick Model?

The New World Kirkpatrick Model is an updated version from Kirkpatrick Partners that keeps the four levels and adds practical guidance for each. It emphasizes planning from Level 4 first and stresses that results depend on what happens after training, not just in it.

Some of the additions are especially useful. At Level 2, it adds confidence and commitment alongside knowledge and skill, because people who know how but don't feel ready rarely change. At Level 3, it adds "required drivers": the coaching, reminders, recognition and accountability that help new behaviors stick. At Level 4, it adds leading indicators, which are short-term signs that you're on track toward the bigger result.

What Should You Ask at Each Level?

At each level, ask questions that point toward the next one. A Level 1 survey that only asks about enjoyment is a dead end, while one that asks about relevance and intent to apply gives you an early hint about Level 3.

Sample Questions for Levels 1 and 2

• "How relevant was this training to the work you do each week?" (1 to 5 scale)
• "How confident are you that you can use this on the job tomorrow?"
• "What might get in the way of using what you learned?"
• A realistic scenario or demonstration that shows the skill, not a recall question.

That third question is especially valuable. Learners will tell you about missing tools, unsupportive managers and conflicting procedures, which are exactly the barriers that block Level 3.

Sample Questions for Levels 3 and 4

• For managers, a few weeks later: "Which of these behaviors have you seen this person use?" with a short checklist.
• For learners: "Which parts of the training have you used, and which haven't you used yet? Why?"
• For the business owner: "Which metric did we agree to watch, and how has it moved against the baseline?"

Keep surveys short, because five well-chosen questions get more honest answers than twenty generic ones, and busy managers are far more likely to finish them.

What Are the Common Criticisms of the Kirkpatrick Model?

The common criticisms are that the levels don't always connect in a chain, that Level 1 gets too much attention, and that the model doesn't tell you how to isolate training's effect from everything else happening at the same time.

Researchers have pointed out that positive reactions don't reliably predict learning, and learning doesn't guarantee behavior change. That's a fair point, and it's why measuring only Level 1 tells you very little about impact. Others note that the model describes what to measure more than how to measure it.

These criticisms are reasons to use the model carefully, not to drop it. Its biggest strength is giving everyone a shared vocabulary. When a stakeholder asks "did it work?", you can ask back: "At which level?" That question alone improves most evaluation conversations.

How Does the Kirkpatrick Model Relate to ROI?

The Kirkpatrick model stops at business results, while return on investment (ROI) goes one step further and compares the money value of those results with the cost of the training. Jack Phillips added that step as a fifth level in what's now called the Phillips ROI Methodology.

You need solid Level 4 data before you can calculate ROI. If you can show that first-call resolution improved and estimate what each resolved call is worth, you can compare that benefit with what the program cost. Our guide to eLearning ROI walks through the formula step by step with a worked example.

Which Mistakes Should You Avoid?

The mistakes to avoid are measuring only what's easy, measuring too late to have a baseline, and asking about everything at every level. Each one weakens the story your data can tell.

• Stopping at Level 1 and calling it evaluation.
• Writing Level 2 tests that check recall instead of skill.
• Forgetting to record baseline numbers before training starts.
• Leaving managers out of Level 3, even though they see behavior every day.
• Measuring all four levels for every course. Save full evaluation for programs that are costly, high-risk or strategic.

A practical approach many teams use is tiered evaluation. Every course gets Levels 1 and 2, and important programs add Level 3. Only a few major programs each year get the full four levels, and perhaps ROI.

The Kirkpatrick Model Works Best Backward

The Kirkpatrick model is a four-level framework that evaluates training by reaction, learning, behavior and results. It's simple to understand, widely recognized and still useful decades after Donald Kirkpatrick introduced it.

The trick is to use it backward. Start with the business result, define the behaviors that drive it, design learning that builds those behaviors, and make sure learners see the relevance. Then measure each level at the right time, with a baseline in place.

Pick one upcoming program and write its Level 4 target in a single sentence. Everything else in your evaluation plan can grow from that.

← All articles