Design for the person who is tired and has forty minutes
Online learning is usually designed as though the learner arrived rested and undistracted. Almost nobody does, and that single fact should determine most of the design.
Picture the actual conditions. It is a Tuesday evening or a snatched hour between shifts. The learner has already worked a full day. They are on a laptop with several other tabs open, or on a phone. Somebody may interrupt them. They have between twenty and fifty minutes before something else takes priority, and they are not certain they will get the same window tomorrow. That is the environment nearly all online learning is actually consumed in, and it is almost never the environment it was designed for.
The designed-for learner is different: they have a clear block of time, they will work through the module in order, they will engage with the reflective activity, and they will return to the discussion forum later in the week. Design decisions follow from that imagined person. Modules run an hour because an hour is a sensible unit. Content is sequenced on the assumption of continuity. Activities assume the learner will be back.
Almost every characteristic failure of online learning follows from the gap between those two people. Long modules are abandoned partway and the resumption mechanism becomes the most important feature in the course. Sequences that require continuity break when there is a week between sessions. Discussion forums, which need a critical mass of people present at the same time, sit empty and are experienced by everybody as an obligation nobody else is meeting.
Designing for the real learner produces different choices, and most of them are subtractive. Segments short enough to be completed in a single sitting, so that stopping is not losing progress. A visible position indicator, so that returning after a fortnight does not require reconstructing where you were. Each segment self-contained enough that a gap does not break it. Assessment that can be attempted without a prior activity that depended on other people being present.
The second thing that follows is about the cost of interaction. Interactive elements are expensive to build and are justified by engagement, which is measured by clicks. A click is not attention, and a great deal of what is built as interaction is a page turn wearing a costume: drag the label onto the diagram, click each icon to reveal the text, complete the sequence to continue. These consume the learner's limited window on an activity that conveys nothing the text would not have conveyed faster.
The interactions that earn their cost are the ones where the learner does something with a consequence: makes a decision and sees what follows, attempts a task and gets feedback on their specific attempt, or practises the actual thing the course exists to teach. Those are worth building and they are much rarer than the click-to-reveal variety, because they require somebody to model the underlying task rather than to decorate the content.
The third is that the completion problem is mostly a design problem rather than a motivation problem. Learners are routinely described as lacking discipline when they do not finish, and the same learners complete things that fit their circumstances. A course that can only be completed by somebody with an uninterrupted evening has selected its audience, and the people it excludes are disproportionately the ones with caring responsibilities, shift work, or a second job.
There is a final point about who these decisions are actually made by, since it explains why the same faults recur. Course structure is usually settled by whoever commissioned the project, on the basis of a scope and a budget, before any designer is engaged. Module length, the presence of a discussion forum, the use of video and the assessment approach are frequently fixed at that stage by people describing what they expect a course to look like. A designer arriving afterwards can improve the execution and cannot change the shape, which is where most of the effect lives.