The Kirkpatrick Model, Explained

Here’s a pattern you’ve probably seen. A team builds a training, runs it, and then measures success by two things: did people show up, and did they like it. Survey scores come back at 4.6 out of 5. Everyone calls it a win and moves on.

The problem is that neither of those things tells you whether the training actually did anything. People can enjoy a session, remember nothing from it, and never change how they work. Donald Kirkpatrick noticed this gap back in 1959 and built a model to fix it. He broke evaluation into four levels, each one answering a different question about whether training actually worked.

This matters more in enablement and customer education than almost anywhere else, because you’re usually not just trying to inform people. You’re trying to get a rep to pitch differently, a customer to self-serve instead of filing a ticket, a partner to stop making the same mistake. If you only measure whether people liked the training, you’ll never know if any of that actually happened.

Liking a training and learning from it are two different things, and learning from it and using it are two different things again.

The model has four levels, and they get harder to measure as you go up, which is exactly why most teams stop at the easy ones.

Here’s the uncomfortable truth. Most training programs never check Level 3 or 4. It’s not because teams don’t care, it’s because those levels are harder to measure and usually take longer to show up. A survey takes five minutes. Checking whether behavior actually changed six weeks later takes planning, patience, and usually a manager or a dataset you don’t fully control.

But Level 3 and 4 are where the actual value lives. If a new onboarding course gets great quiz scores but new customers are still filing the same support tickets, the course didn’t work, no matter what Level 1 and 2 said. If reps pass a roleplay with flying colors but still aren’t using the new pitch on real calls, that’s a Level 3 failure hiding behind a Level 2 success.

This is also where enablement and customer education teams can prove their worth. Anyone can show a completion rate. Showing that a program actually reduced ramp time or cut ticket volume is a different conversation, and it’s the one that gets you a seat at the table.

The hardest levels to measure are usually the ones that matter most.

Here’s where the model gets practical. Most people build training by starting at Level 1: write the content, run the session, then figure out how to measure it afterward. Jim and Wendy Kirkpatrick, who updated the original model into what’s called the New World Kirkpatrick Model, flipped that order on purpose.

Their advice is to start at Level 4. Before you write a single lesson, ask: what business result am I actually trying to move? Fewer tickets? Faster time to first value? Higher deal close rate? Once you know that, you work backward. What behavior change (Level 3) would drive that result? What knowledge or skill (Level 2) does someone need to make that behavior change? And only then, what content and format (Level 1) will get them there?

This single flip changes how you design everything. Instead of asking “what do I need to teach,” you start asking “what needs to be true on the other side of this training for it to be worth building.” That’s a much better question, and it’s the one that keeps training from becoming content for content’s sake.

Design from the result backward, not from the content forward.