How to Measure Training That Sticks: A Kirkpatrick Model Primer for L&D Teams

Ask most training teams how they measure success, and you'll get a similar answer: they send a survey after the session. These surveys usually ask if people enjoyed the training, found it useful, or would recommend it. This is the easiest data to collect, but on its own, it is almost meaningless.

The Kirkpatrick Model, developed by Don Kirkpatrick in the 1950s, is still the industry standard for training evaluation and explains exactly why. It breaks down training impact into four levels, and it's brutally clear about which ones actually matter to the business.

Level 1 - Reaction: Did learners find it relevant and engaging?

This is your post-session survey. It asks questions like: Did people like it? Was the pace okay? Would they recommend it? This data is a useful pulse check; if scores plummet, you know something is broken. Good scores, however, tell you nothing about whether anyone learned anything. People can enjoy a session and retain almost none of the information.

Level 2 - Learning: Did knowledge or skills change?

Here, you start to measure something tangible. Has a pre/post assessment shown improvement? Can the learner describe or demonstrate the new skill after training? Most training teams that go beyond Level 1 stop here. While it's an improvement, it is an incomplete picture because knowing something for a quiz and using it on the job are two different things.

Level 3 - Behaviour: Are new skills applied on the job?

This is where many training programmes quietly fail without anyone noticing. A learner can ace a knowledge check but go right back to their old habits once the session ends. This often happens because nothing in their work environment reinforces the new behaviour. To measure this level, you need to look beyond the training event. This could involve checking usage data, getting manager observations, conducting follow-up check-ins, or using in-product analytics to see if the taught workflow is being used.

Level 4 - Results: Outcomes achieved?

This is the top of the pyramid. It connects training back to the metrics leadership cares about: reduced support ticket volume, higher feature adoption, faster time-to-value, improved retention, and fewer escalations. This is the hardest level to measure cleanly, as business outcomes have many causes and training is just one factor. But it is the only level that proves the training was a worthwhile investment.

Why most teams get stuck at Level 1

Teams often get stuck at Level 1 for a practical reason. Level 1 data is easy to get from a form you are already sending. Measuring Levels 2 to 4 requires dedicated tools: pre/post assessments, usage tracking linked to attendees, and a way to connect training data with business metrics from other systems.

The pyramid shape is meaningful. Each level is harder to reach than the one below it, and each provides information the lower levels cannot. A programme that only reports Level 1 satisfaction scores is flying blind on whether it truly works.

How to advance beyond Level 1

You don't have to build a Level 4 attribution model overnight. Here is a sensible path forward:

Add a simple pre/post knowledge check to your most important sessions. This alone gets you to Level 2.
Pick one behaviour you can observe in-product (like a feature being used) and track it for attendees versus non-attendees. That's Level 3.
Once you have Level 3 data, correlate it with a business metric you already track. Support volume or churn are good places to start building a Level 4 case.

Reaction data tells you if people liked the training session. Results data tells you if the training was worth running at all. Most teams are optimising for the wrong one.