The 90% completion problem

The numbers side by side

Compliance training completes at more than twice the rate of voluntary e-learning. On a standard L&D dashboard, it is the best-performing category in the business.

It is also the category where industry research repeatedly notes that high completion "rarely produces measurable behaviour change."

Why the numbers diverge

Completion measures administration, not attention.

Compliance training completes at 90%+ because it is compulsory, tracked by HR, escalated to line managers, and frequently tied to system access or annual objectives. None of those mechanisms have anything to do with whether the content is any good. They guarantee that the module is finished. They say nothing about whether anyone was present.

Voluntary e-learning completes at 30–40% precisely because nothing forces it. That number is informative — it is a genuine signal about whether people found the content worth their time. The lower number carries more information than the higher one.

What "clicking through" actually looks like

Anyone who has instrumented a compliance module recognises the pattern:

• Sessions where total elapsed time is under half the content's runtime
• Modules completed with the tab in the background for most of the session
• Sections advanced faster than the narration could physically have played
• Assessment attempts clustering at the pass mark, with retries until it clears
• Consistent, sharp drop-offs at the same content position across every cohort

The last one is the interesting one, because it is diagnostic rather than merely damning. When 60% of learners disengage at the same point in a module, that is not a workforce problem. That is a content problem with a timestamp on it.

Why this matters more than a low completion rate

A low completion rate is visible. Someone notices, and something happens.

A 94% completion rate on training nobody absorbed is invisible, and it is worse in three ways:

It creates false assurance. The organisation believes its people understand the anti-bribery policy because 94% completed the module. That belief is load-bearing in a regulatory conversation, and it may be unfounded.

It hides the failure from the people who could fix it. The content team sees green. Nobody rewrites a module that reports 94%.

It wastes the spend twice. You paid for the licences and you paid for the hours. If 40% of a 3,000-person workforce was not attending, you bought 3,000 hours to deliver content to 1,800 people — and the 1,200 hours were charged to the business at full rate.

What to measure instead

Keep collecting completion. Regulators want evidence of delivery, and you need it for audit. Just stop presenting it as an effectiveness measure.

Add three things:

1. Content-level engagement. Which sections hold attention, which lose it, and where sessions end. This is the only measure that tells you what to change. It requires instrumenting the content rather than the LMS, using ordinary browser signals — visibility state, window focus, dwell per section, scroll behaviour. No camera required, and therefore no EU AI Act exposure.

2. Assessment delta with delayed retention. Not the post-test score, which mostly measures short-term recall. The pre/post difference, re-checked at 30 days. Confidence-weighted scoring is worth the extra design effort here, because confidently-held wrong answers are the most dangerous outcome in compliance training and a standard test cannot see them.

3. Domain incident rate. If the training covers data handling, track data handling incidents. If it covers safety, track near-misses. This is the closest thing to ground truth available, and it is usually already being collected by someone else in the building.

The reframe worth taking to your exec team

Try this comparison:

"Our compliance module completes at 94%. Instrumenting it showed that 38% of learners had the tab in the background for more than half the session, and 61% disengaged during the escalation procedure section. We rewrote that section. Engagement through it went from 39% to 74%, and assessment scores on the related questions rose 22 points."

That paragraph does something the completion rate cannot. It identifies a specific failure, attributes it to a specific piece of content, describes a specific intervention, and reports a specific result.

It also makes the case for the measurement itself. Nobody funds an engagement analytics programme on the argument that the current metric is philosophically unsatisfying. They fund it when you show them a number they believed was fine, and demonstrate that it was not.

Frequently asked questions