Let's begin with a moment of honesty. Somewhere in your organisation right now, someone has your annual compliance module open in one browser tab and their actual job in another. The module is "playing". The learner is elsewhere. At the end, they'll guess their way through five multiple-choice questions written to be unfailable, collect their certificate, and the LMS will record a success story.
Multiply that by every organisation on earth and you get a market estimated at US$4.68 billion a year, even back in 2017. The question almost nobody asks out loud: does any of it work?
The research is in, and it's awkward
A Harvard Business School study got the kind of access researchers dream about: five years of internal records from a Fortune 500 multinational, including 2,237 misconduct allegations and roughly 339,000 compliance training participations. The findings deserve to be framed and hung in every L&D office:
- On average, compliance training had no detectable effect on actual misconduct.
- In-person training did reduce misconduct, but the effect faded after about two months.
- Video-based training, the click-through kind that dominates the industry, showed no detectable effect at all. Approximately 19,200 employees needed to be trained in person to prevent a single case of misconduct.
Meanwhile, in a Deloitte survey cited in the same research, half of compliance officers said they measure their programme's effectiveness by training completion rates. Which is a bit like measuring a fire alarm by how many people have walked past it. Harvard Business Review's verdict on the genre was titled, without much diplomacy, "Why Compliance Programs Fail": organisations measure inputs instead of behaviour, and employees resent box-ticking exercises designed to protect the company from them rather than help them do their jobs.
Why the genre fails
It's not because compliance content doesn't matter. Health and safety, privacy, conduct: this is the content with actual stakes. The failure is in how it's made, and the causes will look familiar:
It's written by the policy, for the policy. Most compliance modules are the policy document with a Next button. The legal language survives intact because nobody with editing authority was budgeted to touch it, and legal language is precisely the dialect that adult readers' eyes refuse to absorb.
It answers no question the learner has. Compliance modules are organised around what the organisation must be able to say it told people, not what people need in the moment of risk: what does dodgy actually look like in my role, and what exactly do I do when I see it?
It's designed for completion, not retention. The forgetting curve is brutal and well-replicated: without reinforcement, most new information fades within days. An annual one-shot module is a design that the science of memory predicts will fail, scheduled to fail annually.
The assessment insults everyone. Questions written so nobody can fail also ensure nobody has to think. Learners notice. Thinking is the part that was supposed to be the training.
What actually works
The fix is not more interactivity sprinkled on the same content. It's different content, differently scheduled.
1. Rewrite before you build. Take the policy and translate it into plain language organised around the learner's decisions. This is the highest-leverage hour in the whole project: the FedEx ground operations crew found that rewriting their manuals in plain language meant staff found the right answer 27% more often, in less time, saving US$400,000 in the first year. Same knowledge. Different writing. That's the gap between a module that informs and one that decorates.
2. Teach scenarios, not clauses. The Harvard data showed in-person training worked (briefly) while videos didn't. The plausible ingredient is that rooms force engagement with situations. You can get much of that on screen, but only with genuine dilemmas: ambiguous, role-specific, none-of-these-options-feel-great scenarios where the learner must commit to a choice and see the consequence. One honest dilemma outperforms forty flip cards of definitions.
3. Space it, because memory demands it. That two-month decay in the Harvard study matches the memory research precisely, and the countermeasure is the most replicated finding in learning science: spaced practice. A meta-analysis covering more than 14,000 participants found spaced study beat massed study 47.3% to 36.7% on recall, with the same total study time. Replace the annual 45-minute slab with brief refreshers across the year: a two-minute scenario each quarter beats a yearly module the curve has already eaten.
4. Test for retrieval, not recognition. Testing is itself a learning intervention, more powerful than re-reading for long-term retention. But only if the question makes people retrieve: "A supplier offers your team tickets the week before contract renewal. What do you do first?" forces memory and judgement. "Is bribery (a) good or (b) bad" forces nothing except mild despair.
5. Measure something real. Reported near-misses, policy queries, speak-up rates, audit findings. If completion percentage is your only metric, you're managing the appearance of compliance, and the Harvard study is a 100,000-employee demonstration of how that ends.
The reframe
Compliance training is the only training with a guaranteed audience and genuine stakes, and the industry has spent billions making it the genre everyone opens in a background tab. The opportunity hiding in that failure is real: write it like it matters, schedule it like memory works, and measure it like behaviour is the point.
That's the version we build. Certificates optional. Reading, mandatory in the only sense that ever worked: people actually wanting to.