The Kirkpatrick Model in 2026: What It Gets Right and What It Misses

1. The Four Levels: What Each Measures and What Each Cannot

The Kirkpatrick Model defines training evaluation as a four-level hierarchy, where each successive level produces more valuable evidence of training effectiveness, and requires more effort to obtain. The framework’s elegance is its clarity. Its practical problem is that the effort required increases sharply between Level 2 and Level 3, and most organisations stop at Level 2.

Level	What It Measures	How Collected	Business Value
1 — Reaction	Learner satisfaction, perceived relevance, engagement	Post-training survey (smile sheet)	Low — predicts completion but not behaviour change
2 — Learning	Knowledge gain, skill acquisition, confidence increase	Pre/post assessment, quiz, simulation performance	Medium — confirms learning occurred but not application
3 — Behaviour	On-the-job application of trained skills and knowledge	Manager observation, performance review, call monitoring, operational audit, at 3–6 months	High — first level where behaviour change is confirmed
4 — Results	Business outcomes the training was designed to improve	Operational metrics in CRM, ERP, safety, quality, or financial systems	Highest — directly answers whether the investment delivered value

---

2. The Cascade Break — Why Most Organisations Never Reach Level 3

The cascade from Level 1 through Level 4 is the model’s design logic — each level’s evidence builds on the previous. In practice, most organisations experience what might be called the cascade break: the four levels are treated as four disconnected measurement events because the data infrastructure, the cross-functional relationships, and the time required for Levels 3 and 4 are not in place when the programme launches.

1. Level 1 and 2 data is automatic. Level 3 and 4 data is not. LMS platforms automatically generate completion records and assessment scores. Post-training surveys are sent automatically. Level 1 and 2 data arrives without deliberate effort. Level 3 requires a 90-day follow-up process with manager observations or performance data access. Level 4 requires relationships with the teams who own the operational metrics the training was designed to move. Neither exists without deliberate pre-programme design.
2. The baseline must be established before the programme launches. Level 4 measurement requires a pre-programme baseline of the business metric the training is designed to improve. If that baseline is not collected before the programme starts, the post-training measurement has no comparison point. This requires designing the measurement framework before designing the content — which most L&D commissions do not require.

---

3. The New World Kirkpatrick Model — What Changed and Why It Matters

Jim and Wendy Kirkpatrick’s New World model makes two significant updates to the original that are directly relevant to enterprise L&D practice in 2026. First, it explicitly mandates designing from Level 4 backward, starting with the specific business outcome and working back to the behaviour and learning required to achieve it. This is a design discipline, not just an evaluation discipline. Second, it introduces the concept of required drivers, the management reinforcement behaviours, process supports, and environmental conditions that are necessary for training-produced behaviour change to persist after the learning event.

---

4. What the Model Misses — Three Legitimate Criticisms

1. It assumes a causal chain that is often disrupted. The Kirkpatrick cascade assumes that positive Level 2 data produces positive Level 3 data, which produces positive Level 4 data. In practice, each transition is mediated by factors outside training, manager behaviour, process design, tool availability, organisational culture. Training that produces excellent Level 2 results can produce poor Level 3 results if the environment does not support application. The model identifies this risk through required drivers but does not provide a framework for diagnosing or addressing the environmental barriers specifically.
2. It does not address the forgetting curve. The original Kirkpatrick model is silent on the decay of trained behaviour without reinforcement. A programme that produces strong Level 3 data at 30 days but produces poor data at 90 days; because no reinforcement was built in, has produced transient behaviour change, not durable capability development. This is the gap where the forgetting curve intersects with the Kirkpatrick framework, and it requires a reinforcement architecture that the model’s design does not mandate.
3. It is most commonly applied at its least valuable levels. The structural incentive of the model; that Levels 1 and 2 are easy and Levels 3 and 4 are hard, produces an evaluation culture where organisations consistently measure the two least valuable levels and defer the two most valuable ones. This is not a model failure. It is an implementation failure that the model cannot self-correct for. Addressing it requires a cultural and commissioning change — that L&D functions are expected to demonstrate Level 3 and Level 4 evidence, not just Level 1 and Level 2 completion.

---

In Summary

The Kirkpatrick Model is not broken. It remains the most accessible and most widely understood framework for connecting training to business outcomes. Its logic, Reaction to Learning to Behaviour to Results; is correct, useful, and sufficiently flexible to serve any training context in any sector. The problem is not the model. It is the consistent failure to implement its most valuable levels.

The practical improvement available to most enterprise L&D functions is not a different evaluation framework. It is the discipline to establish Level 4 baselines before programmes launch, build the cross-functional relationships to access Level 3 and Level 4 data, and treat the measurement framework as a design deliverable rather than a retrospective reporting exercise. That improvement is achievable within the Kirkpatrick framework; and it would change the quality of the evidence most L&D functions can provide for their investment significantly.

---

---