Every L&D team can name the Kirkpatrick model. Four levels, introduced by Donald Kirkpatrick in 1959 and still the default language for evaluating training: Reaction, Learning, Behavior, Results. It is taught in every certification and cited in every proposal. It is also, in most organizations, where measurement stops being real. The model is not the problem. What happens to it in practice is.
The four levels, honestly
Level one is Reaction: did people like it. You measure it with the survey handed out as the room empties, and it tells you whether the coffee was good and the facilitator was likeable. Level two is Learning: did they know more after than before, which means testing the same people before the training, not only after. Level three is Behavior: did anything they do on the job actually change, which you cannot see for sixty to ninety days and cannot see at all unless you defined the behavior first. Level four is Results: did a business number move — productivity, error rate, retention, revenue — in a way you can connect back to the training.
Jack Phillips later added a fifth level, ROI, which takes the business result from level four, puts a dollar value on it, and sets it against what the training cost: benefits minus cost, divided by cost. It is the level executives actually asked for, and the one almost nobody produces, because it depends on every level beneath it being done first.
Where it breaks
The pattern is consistent across almost every team. Level one gets done every time, because it is free and immediate. Level two gets done sometimes. Level three and level four — the only two that tell you whether the training was worth doing — get done rarely and badly, and that is where L&D loses its credibility with the business. The reason is not laziness. It is sequence.
Behavior and results cannot be measured after the fact, because both need a baseline that no longer exists once the training has happened. To know whether error rates dropped you need last quarter’s error rate, the specific errors the training targeted, and a way to attribute the change. None of that can be reconstructed the week a stakeholder asks for the impact number. It had to be set up before the course existed, which means measurement is not the last step of a project. It is the first.
Measurement is a design decision, not a report
This is the shift that separates a training team from a performance one. You do not build a course and then wonder how to measure it. You start from the business result you are trying to move, name the on-the-job behavior that would move it, and design the training backward from that behavior — so the assessment tests the behavior and the metric was baselined before anyone opened an authoring tool. Measurement stops being a report you run at the end and becomes the specification you build from at the start.
It also reframes the numbers teams already have. A completion rate is a level-zero metric: it says the course was consumed, not that anyone learned or changed. The same is true of sheer training volume — a library that doubles in size moves no level that matters. More activity is not more evidence. The levels exist precisely to stop confident-looking numbers from standing in for the ones that count.
What to actually do
The honest version of this is not measuring everything. It is choosing the highest level you will actually act on and building for that one. Most programs should aim at level three: a specific behavior, baselined before launch, checked sixty to ninety days after, tied to the one business number the program exists to move. That is enough to defend the work and to improve it. Real ROI, level five, is worth the effort only when the number will change a decision — a program up for renewal, a budget under scrutiny — not as a routine ritual on every course.
None of this requires a new model. Kirkpatrick has been right for more than sixty years. It requires treating the top two levels as design inputs rather than reporting afterthoughts, and being willing to say, at intake, what would count as this program having worked. Teams that do this stop guessing whether training helped. Teams that do not keep sending smile sheets upstairs and calling it evidence.
Measurement is not the paperwork that follows the work. It is the decision that shapes it, made at the start, when you still have the baseline and the honesty to name what the training is actually for. Run it at the end, and the best you will ever have is a room that liked the coffee.