A training program's post-session survey shows 4.6 out of 5 satisfaction, glowing comments, and a strong recommendation rate. This is reported internally as evidence the training worked. It's evidence the training was well-received — a real and not meaningless thing to know — but it is Kirkpatrick's Level 1 data, and Level 1 satisfaction has a well-documented, often weak relationship with Level 3 behavior change, the level that actually determines whether the training changed anything about how people do their jobs.
What the four levels actually distinguish
Donald Kirkpatrick's model defines four progressively more demanding evaluation levels: Level 1 (reaction — did participants find the training engaging and valuable), Level 2 (learning — did participants actually acquire the intended knowledge or skill), Level 3 (behavior — did participants change what they actually do on the job afterward), and Level 4 (results — did that behavior change produce a measurable business outcome). Each level answers a genuinely different question, and a strong result at one level doesn't imply a strong result at the next — the model was specifically built around this distinction, not as a simple ascending checklist.
Level 1 is measured almost universally because it's the easiest, not because it's the most informative
A satisfaction survey can be distributed and analyzed within days of a training session, requires no follow-up, and produces a clean, easily reportable number. Measuring Level 3 behavior change requires observing or surveying participants weeks or months later, ideally with some comparison to their pre-training behavior, which is more expensive, slower, and organizationally harder to arrange. This asymmetry in measurement difficulty, not any genuine confidence that Level 1 predicts Level 3, is why Level 1 data dominates most organizations' training evaluation, often as the only data collected at all.
Participants can rate a session highly while it changes nothing about their actual behavior
An engaging, well-delivered, enjoyable training session can still fail to produce lasting behavior change for reasons entirely separate from how well it was received — insufficient reinforcement afterward, no accountability structure requiring the new behavior be applied, a work environment that doesn't support or reward the change. A participant can genuinely enjoy a session, rate it highly, and simply go back to their prior way of working the following week, and Level 1 data has no way to distinguish this outcome from genuine behavior change, because it was never designed to measure behavior at all.
What measuring Levels 2 through 4 actually requires
Level 2 (learning) requires some form of assessment — a knowledge check, a skills demonstration — administered close to the training itself, distinct from asking whether participants enjoyed the session. Level 3 (behavior) requires observing or gathering evidence about actual on-the-job behavior at a meaningful delay after training, often through manager observation, self-report tied to specific behavioral indicators, or performance data. Level 4 (results) requires connecting that behavior change to a business metric the organization already tracks, which requires the behavior change to be specific and measurable enough to link causally, not just generally described.
What this means for how training gets evaluated
- Report Level 1 satisfaction data as exactly what it is — evidence of a well-received session — not as evidence of effectiveness more broadly
- Build a Level 3 behavior-change check into the training design from the start, not as an afterthought considered once the session is already delivered
- Be skeptical of any "training effectiveness" claim supported only by post-session survey data
- Where the training investment is significant enough to warrant it, connect Level 3 behavior evidence to a Level 4 business result the organization already measures
None of this means satisfaction data is worthless — participants disengaged enough to rate a session poorly are unlikely to apply anything from it. It means satisfaction is a floor, not a proxy, for the behavior change that training is actually meant to produce.