Course completion systems: finishing is an operations problem.

You cannot improve a number two people define differently.

Completion is the most casually used word in online education. Before any of the work below is worth doing, decide precisely what you are counting — because every intervention that follows gets judged against it, and a moving definition can make a course look like it improved when nothing changed except the denominator.

Seven definitions cover it. Write them down once, put them somewhere the whole team can see, and use the same words every time afterwards. This is an afternoon of work with no engineering in it and it is the reason next quarter’s comparison will mean something.

On benchmarks: do not

Somewhere in the research for this you will find a headline completion figure for online courses, and it will be tempting to hold yourself against it. Do not. Those numbers are averaged across free courses, compliance training, university-adjacent programmes and paid cohorts, over populations that did not pay what your students paid and did not arrive the way your students arrive — and almost none of them state which denominator they used, which alone can move a rate by a factor of two.

The comparison that is actually available to you is better anyway: your own last cohort, measured the same way, before and after one change. That tells you whether the change worked. An industry average cannot tell you anything, because you cannot act on the gap.

Session one has to end with something finished.

If there is one change on this page worth making before all the others, it is this. The student decides whether this course is going to be part of their life at the end of the first sitting, and they decide it on evidence — whether anything actually happened.

Most courses open with a welcome video, a tour of the platform and an overview of the modules ahead. That opening is written for the author, who knows how much is coming and wants to frame it. For the student it produces a first session in which they watched some things, learned that there is a lot to do, and finished nothing. There is nothing to tell anyone about, nothing to look at, and no evidence that the purchase is going to work out.

Replace it with a first unit that ends in an artefact. A completed worksheet, a decision made and written down, a setting changed, a first draft, a template filled in, a calculation done with their own numbers. Small is fine. Real is not optional.

What counts as a first win

• →It is complete on its own. Not step one of a five-part exercise. Something that has an end, so that stopping there feels like finishing rather than pausing.
• →It fits the time they actually have. You can find this out by asking, and the answer is almost always shorter than the sitting you designed. Ask on the enrolment form; it is one question and it changes the whole build.
• →It needs nothing installed first. Every account, download or configuration in front of the first win is a gate, and gates are where first sessions end. Move the setup into an optional unit that comes after the student has a reason to do it.
• →It produces something they could show somebody. The showable artefact is what converts a private purchase into a public commitment, and public commitments come back on day two.
• →It is theirs, not yours. A worked example you demonstrate is teaching. The same exercise done with their own situation in it is a win. The difference costs you nothing and changes everything about what happens next.

Curriculum logic and motivation logic disagree here

The shortest complete unit in your programme is rarely the one that belongs first logically. Foundations, definitions and context want to go at the front, and they are genuinely load-bearing later. Put them second anyway. A student who has finished something will sit through the foundations; a student who has only sat through the foundations frequently never gets to the part that would have shown them why the foundations mattered.

Reordering costs a few hours in the outline and nothing in production. Of everything on this page, it is the change with the best ratio of effort to effect, and it is the only one that improves the experience of students who have not enrolled yet.

The unit of design is the sitting, not the module.

Courses are usually built in modules because modules are how the material organises itself. Students do not experience modules. They experience sittings: the amount they can get through between opening the thing and being interrupted.

Size every unit so that a normal interruption arrives after an ending rather than in the middle of one. A student who stops halfway through a forty-minute lesson has to decide, next time, whether to restart it — and restarting is expensive enough that a meaningful share of them simply do not. Four ten-minute units with four completions behind them produce a different next session than one forty-minute unit at sixty per cent.

Then make every unit resumable. “Continue where you left off” is worth more than a beautifully organised table of contents, because the table of contents asks a student who has been away for nine days to remember something, and the resume button does not.

Name milestones as capabilities

“Module 3 complete” tells the student they have moved through your structure. “You can now price a package and defend it on a call” tells them what they got. The second is the reason they bought, and it is the sentence they will repeat to somebody else — which is both a retention mechanism and, incidentally, your best marketing copy, arrived at honestly.

Three or four milestones is usually right. Thirty is a progress bar with extra steps. A milestone should be far enough apart to feel like an arrival and close enough together that the next one is visible from the last.

A progress display that tells the truth

Progress has to be visible, and it has to be honest. A bar that fills as video plays rewards having a tab open. A bar that fills when work is completed rewards the thing you actually want, and it stays accurate when the student tries to use the skill and discovers whether they have it. Students notice the difference eventually; the ones who notice late feel misled, and they are right.

Where a credential is part of the programme, attach it to a capability milestone rather than to arriving at the end — certificates and assessments covers what makes one worth having. A certificate is a reasonable marker of something real and a poor substitute for structure: bolted onto a course people are not finishing, it changes nothing, because the reason they stopped was never the absence of a reward at the end.

Week one is where the completion rate is decided.

Not because week one is the hardest part of the course, but because it is where the habit either forms or does not. A student who finished something on day one and something else on day three has a course in their life. A student who has opened nothing by day five has a receipt.

The week has six moving parts and one of them is not an email. They are worth listing in order, because the ordering is where most of the value is — an orientation message sent before access has landed, or a nudge that arrives before the student has had a reasonable chance to start, does harm rather than nothing.

Two rows in that table are doing most of the work. Minute zero, because the gap between paying and getting in is dead time that nothing later recovers — it is the only part of the week that is purely a plumbing problem, and it is worth testing by buying your own course on a phone. And the branch on day three, because it is the only row that fires on what the student did rather than on what day it is.

All five sends, with subject lines, the fires-on condition for the branch, and what changes for a cohort, are written out in the student onboarding email sequence.

What each message is allowed to do

1. 01Access does one job. Get them in. No curriculum tour, no bonus offers, no newsletter subscription confirmation. Every additional element in this email is a competing link on the most important click in the whole relationship.
2. 02Orientation sets the shape. How the course works, what a sitting costs, where questions go, who answers them, and the name of the first thing they will finish. This is the message that replaces the welcome video you just deleted.
3. 03The first-win nudge points at one thing. Named, specific, small. “Finish the pricing worksheet in unit two — about twelve minutes” beats any amount of encouragement, because it removes the decision about what to do rather than adding pressure to do it.
4. 04The branch notices. It fires on no completed unit, not on day three, and it stops the moment something completes. This is the difference between a system that responds and a broadcast that happens to be badly timed for half the list.
5. 05Momentum reflects it back. What they finished this week, in their words, with the next milestone named as a capability. This is also where the pace expectation gets reset for anyone who is behind, without a word of guilt in it.

Inactive is a threshold, and the threshold is yours to set.

Every re-engagement system starts with a definition problem that most of them skip. Four quiet days means nothing on a twelve-week programme and means the student is gone on a seven-day sprint. Publish the pace first; inactivity is a deviation from it, not a number of days.

So the sequence is: decide what pace you expect, tell the student what it is, then define each signal as a departure from that expectation. This has the useful side effect of making the eventual message defensible — you are not writing to somebody because they have been quiet, you are writing because they are behind a pace they agreed to when they enrolled.

Five signals, in the order they are worth building

1. 01Access granted, nothing opened. The window here is hours, not days. Somebody who bought on Tuesday evening and has not opened anything by Thursday has not made a decision about your course — they have made a decision about their week, and it is still reversible. Wait a fortnight and you are writing to somebody who has already filed the purchase under things that did not happen.
2. 02Started a unit and did not finish it. The signal worth having above all the others, because it carries a location. You are not writing to a student who has gone quiet, you are writing to a student who is stuck on the pricing worksheet, and those two messages have nothing in common. Store the unit, not just the timestamp.
3. 03Finished a milestone and stopped. The dangerous one, because it reads as success in every report. Somebody who completed module two and vanished looks identical to somebody making steady progress until you compare the dates. Milestones are natural exit points — the sense of an ending is real, and it is worth an explicit send that opens the next one.
4. 04Behind the pace you published. Only meaningful if you published a pace. “Two units a week, done in six” gives inactivity a definition; without one, four quiet days on a twelve-week programme and four quiet days on a seven-day sprint produce the same alert and only one of them means anything.
5. 05Asked a question and did not come back. Somebody who posted in the forum, emailed support or raised a hand on a call and then went quiet was blocked by something specific and probably still is. This is the smallest of the five segments and the one where a single human reply does the most work.

The second one deserves the emphasis. Knowing where somebody stopped is worth more than knowing that they stopped, by a wide margin, because it turns a generic check-in into a specific piece of help. It is also the signal most likely to be thrown away by a system that records a last-seen date and nothing else, so check what your platform actually stores before you design a message around it.

Where the signal has to live

A behaviour signal is only as useful as the system that can read it at the moment of sending. If lesson progress lives in the course platform and the email lives somewhere else, then either somebody exports a spreadsheet every Monday, or the message goes out on a calendar and pretends to be behavioural. Both are common. Only the first is honest, and neither scales past a couple of hundred students.

This is the point where behaviour-triggered workflows stop being a nice idea and start being the difference between a system that responds and a newsletter about your course. If you are building your first ones, start with the smallest set and add signals only when a message you already send is provably going to the wrong people.

Name the lesson. Do not miss the student.

The default re-engagement email says some version of “we noticed you have not been around — jump back in.” It is written to everyone who stopped, which means it is written to nobody, and it asks for the one thing the student has already decided against doing.

A message that works has four parts and none of them is encouragement. It names where they stopped. It says, honestly, that this is a common place to stop and why — which converts a private failure into a known feature of the material. It gives one small next action with a real time estimate. And it offers a reply that reaches a person, because some stalled students are stuck on something a two-line answer would clear.

What it does not do is guilt. “You have not logged in for two weeks” is a fact the student already knows and a frame that makes returning feel like admitting something. The message should read as though somebody noticed and wanted to help, because that is exactly what happened — you just did the noticing in advance.

The ladder, and where it stops

1. 01The behavioural message. Automated, fires on the signal rather than the calendar, and names the unit. One action, sized honestly. It should read like somebody noticed, because somebody did — you just noticed in advance and wrote it down.
2. 02The human message. If the first one produced nothing, the second should not be another automated send with a new subject line. A short note from a person that asks a question rather than making a request gets replies that a sequence cannot, and the replies are worth more than the re-engagement.
3. 03The standing invitation. Point at the next live checkpoint by name and date. A stuck student needs a place to bring the thing they are stuck on, and a calendar entry is a much smaller commitment than restarting a course alone.
4. 04The honest exit. Two touches without a response, and chasing stops being help. Offer the thing that is actually useful — a pause, an extension, a transfer to the next cohort, or a refund where your policy allows it. It costs you one student and buys back the reputation that a nagging sequence spends.

That last rung matters more than it looks. Every additional chase after the second one trades a small chance of recovering one student against the certainty of teaching everyone else that your emails are noise. Move non-responders to a slower list, keep the door open, and let the next live checkpoint be the thing that brings some of them back.

Where an assistant drafts these at volume, the draft is the useful part and the send is not — a message that names the wrong lesson is worse than no message at all. Keep customer-facing work behind an approval gate so a person signs off before it reaches a student. The broader pattern, for students who have been gone long enough that this is a campaign rather than a touch, is in the re-engagement campaign guide.

A date on the calendar does work that no email does.

A self-paced course has no external clock in it, which is the feature people buy it for and also the reason a great many of them never finish. A scheduled hour puts the clock back without taking the flexibility away.

The mechanism is not the teaching that happens on the call. It is that a date creates a before, and a before creates work. The unit gets finished on Tuesday night because the call is on Wednesday, and the work is the completion. Which means the checkpoint is doing its job even for the students who intended to come and did not.

Put it where people stop, not at the end

The instinct is to schedule a call at the end of the programme, where it becomes a graduation. A graduation reaches the students who were always going to finish. Place the checkpoint the week after the unit with the worst completion curve instead, so the call exists precisely where students are deciding whether to continue, and can be about the thing they are stuck on.

Judge a checkpoint on whether the work arrives before it, not on how many people attend. A thinly attended call whose deadline pulled a dozen units over the line is doing the thing you scheduled it for. If neither the attendance nor the work is moving, the call is in the wrong week — move it, before concluding that live checkpoints do not suit your programme.

A cohort gets this structure by default, which is most of why the format has the reputation it does. A self-paced programme can borrow it: a recurring monthly call, a start-date intake, published pacing. Cohort versus self-paced is the decision itself; this is the part of it worth stealing either way.

The drop-off report is about the curriculum, not about the students.

Read as a student-behaviour report, the unit-level curve tells you that people are not very committed, which is unactionable and also unkind. Read as curriculum feedback, the same chart points at specific stretches of material and tells you what to change.

Plot unit completion in course order and look at the shape. Four shapes cover most courses, and each one has a different cause and a different fix — which is the entire reason the diagnosis is worth doing before the rewrite.

Fix in order of cost

Check the asset first: a video that fails on mobile, a download that 404s, a quiz that will not submit. It is the cheapest possible cause and it is the answer more often than anybody expects, particularly on a unit that used to complete well and now does not. Then length, then sequence, then the content itself. Rewriting a lesson is the most expensive move available and it should be the fourth thing you try, not the first.

Ask the people who left

The curve tells you where. Only students tell you why, and the students who finished cannot — they are, by definition, the ones for whom the difficult unit was not difficult enough to stop. Exit questions to people who stalled are uncomfortable to send and worth more than any other feedback you collect; student feedback loops covers how to ask so the answers are usable.

Seven chapters, and you cannot do them all this month.

Everything above is worth doing and almost nobody has the time to do it at once. This is the order, for a course that already exists and is not being finished — ranked by how much of the outcome each step controls against what it costs to build.

1. 01Write the definitions down. An afternoon, no engineering, and it makes every number after it comparable. Skip this and you will spend the next quarter arguing about whether completion went up.
2. 02Rebuild the first sitting. The highest-leverage change on the list and the only one that improves the outcome for students you have not enrolled yet. Everything else recovers people; this one stops them needing recovery.
3. 03Send the access and orientation emails properly. Two emails, both of which you are probably sending in some form already, both of which usually have the wrong job. The onboarding sequence is written out in full.
4. 04Wire one stall signal, on the worst unit. Not five signals across the whole curriculum. One, on the unit with the ugliest curve, so you learn whether the message does anything before you build the machine that sends nine of them.
5. 05Put a live checkpoint after that unit. A recurring hour with a date on it does work that no email does. Office hours if the questions are individual, a structured group call if the work is shared.
6. 06Rename the milestones. Cheap, fast, and it changes what the progress display means to the person reading it. Do it while you are waiting for the first cohort of data on everything above.
7. 07Set a review cadence for the curve. Once per cohort, or once a quarter for an evergreen programme. The report is only useful if somebody is booked to read it and empowered to change the curriculum in response.