Quick answer: Smile sheets (Level 1) measure enjoyment, not effectiveness. To prove training changed behavior, design a competency assessment scored by self+others (360-style), send pre and post, then apply Welch t-test per dimension. A medium effect size (dโฅ0.5) with p<0.05 is the evidence your client needs.
The question we keep seeing: Two questions keep arriving together. First, from designers of internal programmes: are we evaluating beyond Level 1, and what would Level 3 evidence even look like here? Then, from external trainers facing a sceptical buyer: how do you prove your training works? Testimonials, they note, are increasingly met with a shrug.
Why 80% of Training Evaluation Dies at Level 1
Kirkpatrick's model gives four levels: Reaction โ Learning โ Behavior โ Results. The vast majority of workplace training stops at the post-workshop satisfaction survey because that is what the tool supports. SurveyMonkey gives you averages of 1-5 stars. It doesn't tell you if the person changed how they give feedback.
Designing for Level 3 (Behavior)
- Competency anchors: "Delivers specific, actionable feedback within 24 hours of observing a behavior" โ measurable frequency or quality scale
- Multi-rater: Self + manager + 1 peer. Reduces self-report bias
- Time separation: Pre = 4 weeks before, Post = 8 weeks after (allows transfer to occur)
- Same instrument: Identical dimension definitions in both rounds for valid pairing
The Statistical Proof
When scores from 20+ participants are aggregated per dimension, the platform computes: pre-test mean ยฑ SD, post-test mean ยฑ SD, delta with direction, Cohen's d (effect size), Welch t (significance), df, p-value, and a binary "significant improvement / not significant / insufficient n" verdict. This goes into a PDF the client can file as evidence that the training investment produced measurable behavioral change.