The Kirkpatrick model: four levels for training that drives results

Executive overview

Training events that employees enjoy rarely move the needle on their own. The gap between a well-received session and actual performance change is where most organisations lose their investment.

The Kirkpatrick model — reaction, learning, behavior, results — is a change management framework, not just a training evaluation tool. The New World version flips the sequence: start at level four (results), work back to level three (critical behaviors), then design the learning.

The training event cannot cross the bridge to performance on its own — what happens before and after the classroom determines everything.

Why most training fails

• The false belief: a well-designed training event is powerful enough to create performance improvement on its own.
• Research confirms it isn't — the event sets the table; performance support and accountability deliver the results.
• Most organisations invest heavily in level one (smile sheets) and level two (testing), then assume the work is done.
• Nothing changes because level three — what people actually do back on the job — is left to chance.
• Buying a new leadership program each year when the last one "didn't work" misdiagnoses the problem; execution at level three failed, not the content.

Level four: results

• Results are not just "did it work in the end" — they are ongoing leading indicators of whether performance is heading in the right direction.
• The ultimate organisational goal (mission, bottom line, public safety) is the fixed flag that never changes; it is the same across all programs.
• Leading indicators are the milestones on the way to that flag — different programs may use different markers toward the same destination.
• Level four is the easiest to measure: HR and business units already track engagement, retention, attrition, and performance metrics — borrow those rather than inventing new ones.
• Clarify the organisation's level four before designing any program; without it, there is no GPS address for the work.

Level three: critical behaviors

• Critical behaviors are the few specific actions most likely to produce the level four outcomes — not a sprawling competency model with dozens of items.
• Employees given too many behaviors default to familiar, easy ones; a short list of critical behaviors focuses effort where it matters.
• Example for emerging leaders: hold weekly performance conversations with employees; conduct structured feedback conversations with struggling team members; partner with peers across departments.
• A 90-day survey alone is not level three evaluation — it is a single data point that tells you little and changes nothing.
• During the 89 days before that survey, there must be active performance support, observation, coaching, and accountability or the behaviors will be forgotten.
• Managers and supervisors must be engaged as partners; if they avoid difficult performance conversations, their direct reports won't change either.

Performance support at level three

• Multiple methods work better than any single intervention: observation, peer support groups, communities of practice, role modeling, post-training debriefs.
• Virtual monthly touchpoints with cohort peers and a few seasoned supervisors keep critical behaviors top of mind.
• These sessions surface what is working (share the practice) and who is struggling (provide targeted help) — low cost, high return.
• The "I can't control what happens after the classroom" objection is an excuse; you can't control it, but you can influence it.
• Crossing the bridge means building a working relationship with supervisors and managers before a program launches — earn the right to partner, don't just sell a solution.
• First step: seek first to understand. Talk to managers about their performance challenges, not about learning objectives. Ask, don't pitch.
• Pilot a proof-of-concept program once trust is established; a visible win converts skeptics faster than any business case document.

When to apply the full model

• Reserve the full four-level effort for mission-critical programs — those most likely to impact the organisation's core mission.
• A one-hour Excel webinar does not warrant this treatment.
• The cost-of-not-doing-it argument: both the Challenger and Columbia disasters happened because level three leading indicators were ignored — safety concerns were raised and overruled.
• Peer-to-peer accountability structures reduce the load on supervisors and make level three sustainable without large budget increases.

Level two: learning

• Level two confirms that participants know what they are supposed to do, how to do it, and what is expected of them back on the job.
• This applies beyond formal training: new policies, new software, new procedures all require level two confirmation that people have the knowledge, skills, confidence, and commitment to act.
• Level two sets the table; it does not deliver the meal.

Level one: reaction

• Level one data is worth collecting but should not be the primary measure of success.
• Smile sheets alone provide insufficient truth — responses may not reflect what will actually happen on the job.
• Best practice (TSA example): combine surveys with instructor debriefs, in-room observation, and supervisor feedback — blended evaluation gives a fuller picture.
• Analyse the data and make specific improvements; collecting and storing it without acting on it is wasted effort.
• Level one is useful for detecting when something went seriously wrong in a session; treat it as a signal to investigate, not a final verdict.

Applying the model beyond training

• The Kirkpatrick model is a change management framework applicable any time behaviour change is needed to achieve a result.
• New policies, system rollouts, org restructures — all can be mapped against the four levels to improve implementation.
• The sequence is always the same: define the result, identify the critical behaviors that drive it, support those behaviors, then design the learning that enables them.