Someone buys your course on a Tuesday. They watch lesson one, maybe lesson two, and then you hear nothing. There is no complaint or refund request, just a login that stops appearing in your platform's activity list.
You usually find out weeks later, when the community is quiet or a student admits they never got round to it. By then the moment to help has passed.
78.3%
completed after finishing a lesson in week one (Ruzuku)
17.2%
completed when week one had no finished lesson
12.6%
median completion across 221 open courses (Jordan)
46%
of paid, verified Harvard and MIT students completed
What the data says about the first two weeks
A widely cited dataset on this is Katy Jordan's analysis of enrolment and completion across 221 online courses, published in 2015. Completion ranged from 0.7% to 52.1%, with a median of 12.6%. Those are open courses with no price attached, so treat them as a rough guide to behaviour.
The useful part is where the engagement went. Jordan found that around half of enrolled students never showed up at all, and that the first two weeks of a course appeared to be critical in gaining student engagement. After that, the numbers levelled out.
Money changes the picture. In Harvard and MIT's open courses, 3.13% of all participants completed in 2017-18, against 46% of the verified students. Someone who has paid is more likely to finish.
None of these figures come from coaching programmes or private courses. They show you where to look, and your own number will differ.
| Source | Measure | Figure |
|---|---|---|
| Katy Jordan, 221 open courses (2015) | Median completion | 12.6% |
| Harvard and MIT open courses, 2017-18 | All participants vs verified (paid) students | 3.13% vs 46% |
| Ruzuku, 921,859 student records (2026) | Completed: finished a lesson in week one vs did not | 78.3% vs 17.2% |
| Ruzuku, September 2026 snapshot | Completion: scheduled vs on demand vs open access | 53.4% / 42.9% / 41.9% |
| Ruzuku, 2026 data | Scheduled cohorts of 11 to 25 active students | 65.2% |
None of these come from coaching programmes or private courses. Use them to know where to look.
What week one tells you
Ruzuku, a course platform, publishes completion data from its own customers, and its cohort versus self-paced analysis is the most relevant public data for creator-run courses. Its strongest signal is the first week: 78.3% against 17.2%, a gap of roughly 4.6 to 1.
Ruzuku is careful about what this does and does not show. The numbers are observational. They do not show that a prompt or a scheduled date caused students to start, because people who were always going to finish may simply start sooner. What you get is an early checkpoint to watch, and a reason to treat a silent first week as a signal.
Its own suggestion for week one is modest. Make module one short enough to finish in under 20 minutes, and add an introductory prompt that asks students for a goal. Try it with one group and see what happens.
Check week one yourself
You do not need any extra software for this. Most course platforms have a progress or completion report, and one look at the end of week one is enough.
Day 0
Student joins
Day 7
Not finished lesson one? Write, name the lesson
Day 14
Still at zero? One more note
After
Stop. A third nudge feels like pressure
- 01
Seven days after each student joins, open the progress report for your course and list everyone who has not finished lesson one.
- 02
Write to each of them individually and name the lesson. "Lesson one is the short one on setting your goal, and it takes about ten minutes" gives them a small, specific thing to do. "Just checking in" gives them nothing.
- 03
Ask one thing they could answer in a sentence, such as whether something in the lesson put them off.
- 04
Do the same at the end of week two for anyone still at zero, then stop. A third nudge starts to feel like pressure.
Check what your platform counts as progress before you trust the report. Teachable's course reporting guide says its lesson completion report measures how many lessons each student marked as complete, and that a lesson is only marked complete if the student clicks the Complete and Continue button. A student who watched the whole lesson and closed the tab can still show as zero. Your message will sort that out quickly, and it costs nothing if you got it wrong.
Two reasons students stall
A student can go quiet for two different reasons, and they need different fixes.
The first is that they got stuck. They hit a lesson that did not quite land, had a question that felt too small to email you about, and decided it was easier to stop than to ask. Nobody was there at 9pm on a Sunday, so the question went unanswered and the course went unopened.
The second is that nobody expected them. Ruzuku's September 2026 snapshot of courses on its platform found that scheduled courses had 53.4% recorded completion per enrolment, against 42.9% for on demand courses and 41.9% for open access ones. Its explanation is social accountability: when there is a live session on Tuesday, you show up because other people are expecting you.
Reason 1
They got stuck
A question too small to email, 9pm on a Sunday, nobody there.
Fix: an answer
Reason 2
Nobody expected them
No date, no group, no one who would notice.
Fix: a person, a date, or both
You can fix the first reason with an answer. The second needs a person, a date, or both.
Find the stuck points before students leave
If you have a Second Self beside your course, its conversation log already shows where students get stuck. You do not need a survey, because students ask the question at the moment it occurs to them.
Personify's guidance on reading your conversations gives a way to sort the log. Every "I do not have that information" is a gap with the question already written for you. A conversation with one message means someone asked once and left, and the answer they got is usually the reason. Wrong answers get fixed first. Questions you did not expect show you what you have been assuming about your students.
For a course creator, the log works as lesson feedback you never had to ask for. If the same two or three lessons keep coming up, that is where your students stall, and you can rewrite them or add a short clarifying video.
A worked example: one Monday with your conversation log
The example below is made up. Say you run a course with eight modules and a Second Self sits beside every lesson. Here is how ten minutes on a Monday morning could go.
- 01
Open the conversations list and look at the message count column. Open the rows with a count of two first, which means one question and one reply.
- 02
Read the answer each of those students received. If the reply was vague or a refusal, you have found the likely reason they stopped.
- 03
Note which lesson each question is about. If three of the conversations with one message point at module three, that module is where your students stall.
- 04
Turn the repeated refusals into Q&A sources, written in your own words, so the next student with that question gets a real answer. If a reply was simply wrong, fix it straight away in Test Arena, and if the wrong fact came from an old document, replace that source too.
- 05
Write to the students who went quiet and mention the specific question they asked. The guidance on following up says that is what proves a person read the conversation.
The order matters. Read the quiet students first, fix the answer, then write to the person.
Put the help beside the lesson
Where you put help matters as much as having it. The guidance for using a Second Self inside a course says a widget on the lesson page gets used, and a link in a welcome email gets forgotten by module three. The placement guide makes the same point about single announcements. A single mention to your list gets one spike and then nothing, and persistent placement keeps working. It also lists "inside your course, next to the lesson" as one of the places people already have questions.
A student who has to leave the lesson, find an email from two weeks ago and click a link is already halfway out of the course.
Feed it your lesson transcripts. Transcripts of your own video lessons are ideal, because they carry both the content and your delivery. A student asking "can you explain that again more simply?" gets your explanation in your words.
Write the greeting with the student in mind. The welcome message guidance says to name two or three things it can help with, because that gets a much higher reply rate than a blank invitation, and to say plainly that it is an AI, which costs nothing and buys trust.
Keep it optional
Do not make talking to it compulsory and do not use it to quiz people. The course guidance is direct about this. Your students are adults doing this voluntarily, and being made to check in reads as homework. Keep it as an optional layer beside the material.
A Second Self answers questions. It will not turn up on a Tuesday evening and make someone feel expected.
Do the accountability part yourself
If students work through your course at their own pace and stall, consider what you have put in place that expects them. A start date with a first live call is one option. A deadline for lesson three is another. A weekly note that names the lesson everyone is on suits some groups better.
Ruzuku describes a hybrid model for this: recorded core lessons, plus a live session once a week or every two weeks. Its snapshot does not predict a completion rate for hybrid courses, so you measure your own. If you go live, note that in its 2026 data, scheduled cohorts of 11 to 25 active students recorded 65.2% completion.
Live sessions do not suit every audience. Parents and shift workers often need to learn at midnight or on a Saturday morning, and a fixed Tuesday call excludes them. If that describes your students, a soft weekly note and a deadline for lesson three cost you far less than a live call.
If you want to offer both formats, Ruzuku advises two separate courses rather than one shared course. Housing each format in its own container keeps the cohort together and stops self paced students from receiving reminders for calls they cannot attend.
Pick one structure and keep it small, then test it with your next group.
Where a Second Self fits
A Second Self answers the students who get stuck at 9pm and shows you which lessons cause the questions. It does not replace a live call or you noticing that someone has gone quiet.
If you run your course on Kajabi, Thinkific, or Teachable, the setup guides are at Kajabi, Thinkific, and Teachable. The course creators page covers what it does for students. For the numbers behind course completion, see the online course statistics.
Look at your week one report on Friday and write to everyone still at zero.
Answer the student stuck at 9pm
Put a Second Self beside your lessons, trained on your own transcripts, and read where students get stuck.
Free tier available. No credit card needed.