Reaction, Learning, Behavior, Results—the four-level ladder for proving training worked, and what each level actually requires.
The Kirkpatrick model grades training evaluation in four levels: L1 reaction (did they like it), L2 learning (can they demonstrate it), L3 behavior (do they do it on the job), L4 results (did the business move). Each level needs its own instrument—smile sheets can't do L3's job.
L1 survey · L2 scored pre/post · L3 30-90 day behavior check · L4 KPI comparison vs baseline/control.
A coaching program reports L2 lift +23% and L3 manager-check confirmations at 60 days—evidence, not enthusiasm.
Design evaluation before delivery: pick the deepest level your data access supports, and build its instruments into the program itself.
Levels are a causal chain, not a scoreboard—L4 without L3 is attribution theater. And stop apologizing for L1; bad reactions kill adoption.
No—L1+L2 for most internal programs; add L3 for behavior-critical skills; L4 when finance will look.
ROI (Phillips) monetizes L4 minus costs—powerful, assumption-heavy, and often unnecessary if L4 is credible.
Turn the concept into a live assessment：Kirkpatrick Model
Browse templates →