We looked at start-to-finish data for every path taken on EduStream in 2025. The drop-off is not gradual. It clusters hard in days fourteen to twenty-two, and it is largely independent of subject, price and prior experience.
The pattern is consistent enough that it is worth taking seriously as a design problem rather than a character problem.
The schedule was built for a good week
Nearly everyone plans their study time during a week with unusual amounts of free time, or in the enthusiasm just after buying a course. Two evenings and a Saturday morning feels reasonable on the day you decide it.
Week three is the first ordinary week. Something runs late, someone is ill, a deadline moves. The schedule had no room in it, so the week is a total loss rather than a reduced one. Two lost weeks in a row and the habit is gone.
Plan for the ordinary week instead
Learners who finish paths do one thing differently, and it is visible in the data: their sessions are shorter and more frequent. Four sessions of forty minutes beats two sessions of two hours, by a wide margin, on completion rate.
Shorter sessions survive a bad week. A forty-minute block can be moved. A three-hour Saturday block, once lost, is simply gone.
Give the schedule a floor, not a target
Set the smallest session you will still count as showing up. Fifteen minutes is a common floor and it works, because the purpose is not the fifteen minutes. It is that the streak survives the week, and the following week starts from something rather than nothing.
Where the platform helps and where it does not
Progress bars and streaks are useful up to a point: they make an interrupted path easy to resume, which removes the friction of finding your place again. That is a real effect and it is small.
What actually predicts finishing is the shape of the schedule you set in the first week. We now ask about it during onboarding, and we suggest four short sessions over two long ones. Completion in the affected cohort rose by nine points, which is the largest single improvement we have made to the product without changing a single course.