How to design training that sticks

If you can skip the course, score 100% on the final test, and collect the certificate, you may have proved you already knew the material. What did the course add?

My starting point is that good training should make you work. Mentally, I mean. Nobody needs to do three laps around the desk. The effort I’m after is the moment you have to look again, notice something you would normally miss, or reconsider an answer you were sure about.

Make the decision difficult for the right reason

Suppose your training covers suspicious payment requests. “Understand the policy” gives you a topic. “When a supplier emails new bank details, pause the payment and verify the change through a known contact” gives you something to teach, practise, and check.

Make the request plausible: it comes in an existing email thread, the invoice is due today, and the supplier’s message includes a phone number. Ask learners what they would do before showing the rule. Then ask which number they would call and why. The challenge is spotting that a familiar thread and a plausible explanation still don’t verify the change. The instructions themselves can be perfectly clear.

Give feedback on the reasoning and let people try again. A week later, offer a fresh case with the same underlying risk and fewer cues. In Roediger and Karpicke’s experiments, retrieving material from memory improved delayed recall compared with additional study. That supports making room for recall after a gap. Whether learners can apply the idea at work still needs a separate check.

Give each measurement a job

SCORM is useful when you need a course to run in an LMS and report outcomes such as completion and scores. SCORM 2004 can also report interactions, if the course sends them and the LMS makes them available. Keep that reporting when it serves a real need.

Use xAPI when useful evidence comes from several places: a course, a practice simulator, a follow-up exercise, or a supervisor’s observation. Each instrumented tool can send a structured statement to a Learning Record Store. Agree on what those events mean before collecting them; a pile of “completed” events across three systems is still a pile of completions.

• Course completion: Did the learner finish the assigned training?
• Delayed scenario: Can they choose a safe response to a fresh example without a hint? If many miss the same step, revisit how you teach it.
• Observed application: In a supervised exercise or a suitable real task, did they verify the change using a trusted contact? Record which criterion was demonstrated; don’t send actual bank details or customer data to the learning system.

A simulation result is evidence about the simulation. An observation is evidence about a particular task. Neither tells you, on its own, how someone will behave every time. If a signal won’t change what you teach, what you offer next, or what you can responsibly attest to, leave it out.

Give people a reason to care about the score

For some people, a certificate is the goal. A sought-after AWS certification can be worth weeks of preparation because it carries recognition outside the course. A completion certificate can be useful too, but it makes a narrower promise. Say what it certifies: attendance, a pass on an assessment, or a skill demonstrated in practice.

My bet is that for many learners, the more immediate prize is a good result they can compare with someone else’s. There’s a reason people ask “What did you get?” after a test. We like knowing we did well, and sometimes we like knowing we did better. That’s an instinct a training designer can work with.

That’s also why I’d give a final test a second job. It can check what people know, but it can also give them a result worth trying to beat. The catch is that the score tells people what to practise. If the test only asks them to recognise a sentence from the slides, a leaderboard gives them a reason to get better at recognising sentences from slides. Give points for speed and you’ll get quick answers. Give points for spotting a risky payment request, choosing a safe response, and explaining why, and the competition gives people a reason to practise those decisions.

A plain leaderboard can work when the contest is short, the rules are clear, and people have a fair shot. I’d try an opt-in weekly challenge with a fresh case and a score based on the quality of the decision. Reset the board each round so a newcomer has a reason to join. Keep the required assessment separate, and don’t expose everyone’s results by default.

I’d also offer a private personal best or a harder case unlocked by a good result. Some people want to beat the room; others want to beat their last attempt. A team challenge is another option: score how well the team explains its decision, so helping a colleague is part of winning. Points and badges are useful only if they make one of those goals clearer.

Try it on one course

Pick one decision that matters. Replace a stretch of explanation with a short case and feedback. Bring back a different version a week later. Keep completion reporting, and compare delayed performance with the old approach on equivalent cases. If you add a reward, check whether more people come back to practise and whether their decisions improve. More clicks alone won’t tell you either.

Before adding another slide, event, or badge, ask what you want someone to do differently next Tuesday. Then give them a chance to practise it.