Training Evaluation, ROI, and Continuous Improvement
Levels of evaluation
Kirkpatrick's four levels provide a useful structure: reaction (learner perceptions), learning (knowledge and skill gains), behavior (on-the-job application), and results (business outcomes). Reaction data is easy but weak evidence of impact. Behavior and results require follow-up and often supervisor input. Plan evaluation during design, not after delivery.
Measuring impact and ROI
To estimate ROI, isolate the effect of training where possible (control groups, trend lines, estimates from stakeholders), convert results to monetary value, and compare with fully loaded costs including design time, delivery, materials, travel, and lost productivity. Express ROI as net benefit divided by cost, usually as a percentage. Be transparent about assumptions; questions often test whether you isolate effects before claiming credit.
Continuous improvement
Use evaluation data, observation, and learner input to revise content, methods, and transfer supports. Distinguish formative evaluation (during development, used to improve) from summative evaluation (after, used to judge worth). Track whether the original performance gap closed, not just whether learners enjoyed the session.
Traps
- Treating high satisfaction scores as proof of learning or impact.
- Claiming business results without isolating other causes.
- Omitting hidden costs from ROI calculations.
- Evaluating only once and never revising.
Worked example
A program costs 50,000 and produces an isolated net benefit of 75,000. ROI equals 75,000 divided by 50,000, or 150 percent. State the isolation method and assumptions alongside the figure.