The Training Went Great. So Why Did Nothing Change?

The room had energy. People laughed in the right places. Nobody was on their phone. Then the feedback forms came back: 4.8 out of 5.

Three weeks later, the team works exactly the way it did before.

If you have run, bought or delivered training, you have probably been here. It does not mean you picked the wrong trainer. It means we are measuring the wrong thing. In this article, we look at what research says about measuring training effectiveness, why a great session can teach less than a messy one, and four small checks you can try next week.

Why doesn’t 4.8 out of 5 mean the training worked?

When we judge a learning session, whether as a leader, an HR practitioner or a trainer, we tend to look at two things. The first is how the room felt: energy, laughter, attention. The second is the score on the feedback form at the end of the day.

Both feel like evidence. But one is a feeling, and the other is often shaped by politeness. Nobody wants to give a low score to someone who just spent three hours in a room with them. In Indonesia we have a word for that: sungkan.

It is not a rating, but it fills in the form all the same.

In Kirkpatrick’s well-known four-level model of training evaluation, this score sits at Level 1: reaction (Kirkpatrick & Kirkpatrick, 2006). It tells you whether people liked the session. It does not tell you whether they learned, changed how they work, or moved a business result. Too often, it is also where measurement stops.

Meanwhile, the sessions keep running. People go back to their desks unchanged. And at some point, the next training budget gets harder to approve, because “we didn’t see any impact.”

So what has anyone actually measured?

What does research say about measuring training effectiveness?

Two studies, from two very different worlds, give the same answer.

The first comes from a classroom. Louis Deslauriers and colleagues at Harvard taught the same physics students in two ways. In one class, an instructor gave a polished, clear lecture. In the other, students worked through the problems themselves in small groups, with the instructor guiding them. Then the researchers measured two things: how much students felt they had learned, and how much they actually had.

Students felt they learned more from the polished lecture. But they scored higher on the test after the class where they had to work it out. The feeling of learning and the real learning pointed in opposite directions (Deslauriers et al., 2019).

The feeling of learning and the real learning pointed in opposite directions.

The second comes from the workplace. Brian Blume and colleagues combined the results of 89 studies on training, mostly from Western workplaces, covering 12,496 people. One of their questions was whether enjoying a training predicts using it on the job afterwards. The link was tiny. How much people liked a session explained less than 1% of whether they applied it later (r = .08; Blume et al., 2010).

One study is about students, the other about employees. Both say the same thing: how a session feels is a poor guide to what it changes.

Which leaves one question: why?

Why does the “great” session often teach less?

Psychologist Robert Bjork gave a clear answer back in 1994. He separated two things we tend to mix up: performance and learning.

Performance is what you can see during the session: people nodding, answering correctly, following along. Learning is what stays and shows up later, when the trainer is gone. The two do not always move together.

When a trainer explains something smoothly, participants follow easily. It feels like progress. When participants have to work it out themselves (try, get it wrong, try again), it feels slower and harder. Bjork called these “desirable difficulties”: the kind of struggle that makes learning stick, even though it feels worse in the moment (Bjork, 1994).

So the smooth session looks great on the day. The harder one is more likely to be the session that sticks.

If we only measure the day, we will keep rewarding the wrong one.

Why do people say two different things?

This explains something many of us have heard from the very same participants.

Nobody lied. The first set is about the room. The second is about the job.

A feedback form collected at five in the afternoon can only hear the first set.

What would count as proof?

If applause is not proof, what is? Kirkpatrick’s model already points the way. Above reaction sit three more levels, and each one can be turned into a plain question:

• Can they do it? (Level 2, learning.) Not “did they understand?”, but can they show it, in a role-play, a real case or a piece of work.
• Are they doing it at work? (Level 3, behaviour.) Weeks later, in the real job, with no trainer in the room.
• Did it move something that matters? (Level 4, results.) A number the business already watches: fewer errors, faster onboarding, fewer complaints in one team.

Each question moves the measure away from the room and towards the job. And each one is easier to answer if you decide what proof looks like before the session starts, not after.

What can you try on Monday?

You do not need a new evaluation system to start. Here are four small checks. None of them takes more than a few minutes.

These checks will not give you a perfect number. They give you something more useful than 4.8: a picture of what changed at work.

So what is a 4.8 actually worth?

A high score and a lively room are good signs that people enjoyed themselves. That matters. But it is not the same as change.

So let’s stop reading applause as evidence, and start deciding what we would accept as proof.

This article grew out of our media-partner work for LOU&GE × Learning Design Club’s “Full Stack of Learning” (14 November 2026).

Want to Think This Through for Your Team? Let’s Talk →

---

References

• Bjork, R. A. (1994). Memory and metamemory considerations in the training of human beings. In J. Metcalfe & A. P. Shimamura (Eds.), Metacognition: Knowing about knowing (pp. 185–205). MIT Press.
• Blume, B. D., Ford, J. K., Baldwin, T. T., & Huang, J. L. (2010). Transfer of training: A meta-analytic review. Journal of Management, 36(4), 1065–1105.
• Deslauriers, L., McCarty, L. S., Miller, K., Callaghan, K., & Kestin, G. (2019). Measuring actual learning versus feeling of learning in response to being actively engaged in the classroom. PNAS, 116(39), 19251–19257.
• Kirkpatrick, D. L., & Kirkpatrick, J. D. (2006). Evaluating training programs: The four levels (3rd ed.). Berrett-Koehler.