Ninety-eight per cent completed it. It still didn't work.
A financial services firm's compliance training sat at 98% completion for years while near-miss reporting never moved, so I rebuilt it around the fifteen seconds where the policy runs out and judgement has to take over.
The problem
A financial services firm had run the same annual compliance module for years. Completion sat at 98%, year after year, and everyone treated that as good news.
Nobody looked closely enough at what sat beside it. Near-miss reporting, the harassment and bribery-risk kind, stayed just as flat across the same years. Post-incident reviews kept telling the same story: staff finished the module, passed the quiz, and still froze or misjudged the moment a real grey area put them in the room. A senior colleague's after-hours dinner invitation that tipped into pressure. A supplier gift that arrived during a live tender. The module had never once asked anyone to work through a situation like that. It only asked them to recognise, in the abstract, that the rules existed.
Leadership had seen the flat near-miss line before, twice, and both times the fix was another mandatory module. Completion fatigue went up each time. Judgement never moved. The 98% completion figure was not the healthy number it looked like. It was the reason nobody had gone looking for the culture underneath it.
The learner and the constraints
The people this training has to reach are not new to compliance. Some of them have sat through the same module three years running. They can recite the policy in their sleep. What none of them had ever done was sit inside the fifteen seconds where the policy runs out and a decision still has to get made, with a client in the room and no clause that quite covers it.
Legal, HR and Compliance were part of the audience too, just from the other side of the table. Nothing I built would reach a learner without clearing all three first, and they do not agree by default. A scenario had to work for a nervous new starter and a risk-averse general counsel at the same time, or it never left the review room.
The question I set myself: how do I get people to practise the judgement call itself, not just recognise the policy sitting behind it?
The approach, and why
I did not design a fourth module. I designed one hard scenario and refused to build a fifth thing after it.
- Turn down the obvious fix. Two previous rounds of "add another module" had raised completion fatigue and moved nothing. I treated that as the evidence it was, not a reason to try a third time.
- One scenario, not another broad one. "The Dinner Invitation" puts a mid-level employee through an escalating, ambiguous situation: a client dinner that drifts past normal hospitality, then a supplier gift landing mid-tender. No option in the interface is flagged as correct.
- Build the climb into the branching itself. Each stage sits on its own rung of Bloom's, from restating the policy in your own words, through the live decision, to judging your own response against the firm's values afterwards. The difficulty was designed into the branch map before a line of scenario text existed.
- Negotiate tone with Legal, HR and Compliance as we go, not once at the end. Draft branches went round Design, Develop, Evaluate, repeated, with a change log recording what each function flagged and what changed because of it. That record is the proof the negotiation happened, not just the sign-off.
- Fold the reinforcement into the measurement, not a second deliverable. A spaced, un-taught follow-up scenario sits inside the evaluation plan itself, re-testing the same judgement at a distance, on a different grey area, without re-teaching a single rule.
- Measure whether judgement moved, not whether the module got finished. The evaluation plan is weighted at Kirkpatrick Level 3: Behaviour, tracking near-miss reporting quality and manager-observed judgement calls, not a second completion number. Built to WCAG 2.2 AA throughout, so the branching, the clock and the feedback work for everyone in the room.
Artefacts
The full branching build for "The Dinner Invitation," the stakeholder governance framework with its round-by-round change log, a Kirkpatrick evaluation plan weighted at Level 3, and an illustration prompt set calibrated to a restrained, FCA-appropriate tone. All available to walk through on request.
The outcome
The honest number here is the one the firm walked in with. 98% completion, for years, next to near-miss reporting that never moved. Everything the redesign is built to shift sits in amber until a real cohort has been through it.
What I took from it
Completion was never the evidence of a healthy culture. It was the reason nobody had gone looking for the unhealthy one underneath it. You do not build judgement by explaining the rule one more time. You build it by putting someone inside the exact moment the rule runs out, and letting them find their way through it somewhere that isn't a real client dinner. That is what separates training a firm can point to from training that changes what someone does the next time the invitation lands.
Further reading: compliance training that changes behaviour, measuring behaviour, not completion.