The short version
- Completion rate is the most reported number in corporate learning and one of the least informative.
- A course nobody opens is usually a timing problem wearing the costume of a content problem.
- People reach for help at the moment of the task, not the week before it.
- Report what changed in the work, not what percentage finished the module.
Completion rate is the number most learning teams are asked for first, and it is close to useless as a measure of whether anyone learned anything.
It tells you a module was marked finished. It does not tell you whether the person could do the task afterward, whether they needed to, or whether they clicked through at double speed with the window in the background.
What completion actually measures
Completion measures compliance. That is a legitimate thing to measure, and in a regulated environment it is the thing an auditor asks for. I have built programs where a documented completion record was the entire point, and getting to one hundred percent with zero findings at a federal audit was the deliverable.
The mistake is carrying that number into contexts where nobody is auditing anything, and treating it as evidence of capability. Those are different questions and they need different instruments.
The timing problem hiding inside a content problem
When a course goes unopened, the first instinct is usually to rebuild it. Better video, tighter script, more interaction.
Sometimes that is right. More often the content was fine and the delivery moment was wrong. A forty minute course assigned in advance of a task competes with the task. A two minute job aid at the moment of the task competes with nothing, because the person is already stuck and already looking.
People do not go looking for training. They go looking for the answer, and they go looking at the exact moment they are blocked.
Design for the moment of need
The practical version of this is unglamorous. Break the long course into the specific decisions it is actually preparing someone to make. Put each one where that decision happens: in the tool, next to the form, in the channel where people already ask each other.
Keep the full course for the people who want the whole argument, and for the record when someone needs to prove the program exists. But stop treating it as the primary delivery vehicle for something a person needs in ninety seconds.
A better number to report
Pick something observable in the work itself. Error rates on the thing you trained. Time to complete a task. Support tickets on the specific step people kept getting wrong. Repeat visits to a job aid, which tells you it is being used rather than assigned.
Those numbers are harder to collect than a completion percentage. They are also the only ones that answer the question your sponsor actually asked, which was never “did they finish it.”