Cohort-Based Learning: Meaning, Examples and How to Run One

Go from storyboard to an interactive course in minutes

Start Free

Picture sixteen graduate engineers starting at the same consultancy on the same Monday in February, ten in Brisbane and six in Perth. The firm could give each of them a self-paced induction course and let them work through it alone. Or it could run them as a cohort: one start date, a new week of material every Monday, small groups that review each other's work, a live session every second Thursday and a principal engineer who reads what they post. The second option costs more of a senior person's time. Whether it buys anything depends on what the graduates need to learn, and the completion figures usually quoted in its favour do not answer that.

I'm an instructional designer at Coursebox, where we make AI tools for employee training programmes, so I'll mention our product where I tested it. In this guide I set out what cohort-based learning is, how a cohort differs from a class and from a self-paced course, what the completion evidence actually measured and where cohorts are used at work. Then I plan one graduate intake element by element, from the shared start to the facilitator, and turn it into a week-by-week plan. Along the way there are three real cases, a product test and two free templates.

What is cohort-based learning?

Cohort-based learning is a way of running a course in which a group of learners starts together, works through it on a shared schedule with common deadlines, and learns partly through planned work with each other, usually with a facilitator.

View the full-size diagram

In everyday use, a cohort is a group of people who share a starting point. In learning, the shared starting point is the intake: everyone begins on the same date and moves through the same weeks. Nobody owns the term, and there is no standard definition or originator, so these features describe how cohorts usually work rather than a fixed model. A cohort is a way of delivering a course. It is not an instructional design model in the sense of the frameworks in our guide to instructional design models, and you can use any of those models to design the content a cohort works through.

Five elements are common in cohort programmes. The shared start and shared pace make a group a cohort; peer work, live touchpoints and a facilitator are what make it worth running as one.

What are cohort-based courses?

Online, a cohort-based course is usually a course with an intake: it opens on a set date, everyone works through the weeks together, and enrolment closes until the next intake. The contrast is an on-demand course you can start at any time. Inside organisations the same idea goes by other names, such as a graduate intake, an onboarding class, a leadership cohort or a learning community.

Cohort vs class: what is the difference?

A cohort is a group that stays together across a whole programme, while a class is the group that meets for one course or one timetable slot. In a university, a cohort usually means students who enter a degree together and take most of their units together. In the words of a 2011 review of academic cohorts by Lei and colleagues, cohort education "places students pursuing the same field of education into learning groups that take a majority of their class work together according to a pre-determined schedule of classes." That is what "cohort" means in college and graduate school: your intake year, the people you go through the programme with.

A class can be part of a cohort programme, and a cohort can take many classes. The difference is how long the group lasts and what it shares. A class shares one subject, a teacher and a room or timetable. A cohort shares a start date, a pace and, in a well-run programme, work that its members do with each other.

A cohort is not the same as a live class

Cohorts are often assumed to mean live video sessions. They do not. What defines a cohort is the shared start and the shared deadlines, not meeting at the same moment. A 2019 MIT master's thesis on course pacing describes the scheduled version of an online course this way: "Instructor paced courses follow a set schedule. The course team sets specific due dates for graded assignments and exams, and students complete the course within a defined time period." It adds that this design "fosters a sense of community by encouraging students to progress as a cohort."

A randomised trial with US maths teachers ran its cohort with no live sessions at all: the teachers moved through the course week by week and talked in an online discussion forum. An asynchronous cohort, with a weekly release, a discussion that has a deadline and perhaps one optional call, is a legitimate design. It suits shift workers and people spread across time zones, who may never be free at the same hour.

Cohort vs self-paced learning

In a self-paced course, each learner starts when they like, moves at their own speed and usually works alone. In a cohort, the group starts together and keeps to a shared schedule. Neither is better in general. The comparison below sets out what each format gives and costs; for self-paced design in depth, see our guide to self-paced courses.

The only like-for-like tests I found ran the same course both ways, with learners assigned at random. The clearest is a trial with American maths teachers.

Case study: An online course for US maths teachers, run two ways

Online professional development for teachers was growing fast in the mid-2000s, and the advice to run it as facilitated cohorts rested on observation and surveys. Researchers at Education Development Center and Boston College set out to test it with a scored outcome.

• What they did: They ran one 10-week course on building algebraic thinking in the middle grades in two versions with the same readings, assignments and facilitators. In the facilitated cohort, teachers moved through the sessions together, posted at least twice per session in an online forum and got extended feedback from a facilitator. In the self-paced version, teachers had no contact with each other, wrote private journal entries and got email feedback on assignments. Neither version had live sessions. They randomly assigned 97 teachers, 49 to the cohort and 48 to self-paced.
• Evidence and result: In Carey and colleagues' 2008 report in the Journal of Technology, Learning, and Assessment, more teachers dropped out of the cohort (26) than the self-paced version (18), and 53 completed the course and every measure. The effects on the intended outcomes, including a scored maths test, "generally did not differ significantly between the two versions of the course." Facilitators spent about 11 hours a week on each cohort class and about 3.5 on each self-paced class.
• Limitation: One course, one audience of volunteers, and heavy dropout, so "no significant difference" is not proof that the formats are equal. Apart from the maths test, the outcomes were self-reported.
• Lesson: With strong content and assignments, the cohort was not automatically the active ingredient, and it cost about three times the facilitator hours. Run a cohort when the job needs something only a group can give, and judge it against a well-built self-paced course, not against a weak one.

A second trial points the same way. In Canada, primary care teams in a palliative care programme called CAPACITI took the same three online modules either self-directed or with a weekly live session led by a palliative care clinician. Of 566 people enrolled, 378 finished the first module. Matthew and colleagues reported in 2025 that "No significant differences were observed between the facilitated and self-directed groups across either survey instrument." The outcomes were self-rated, so the trial says nothing about practice with patients.

Two trials with null results cannot show that cohorts make no difference. They do suggest that a shared schedule and a facilitator, on their own, did not produce measurably better results in these settings. The peer work, the deadlines and the facilitator have to do something the content cannot.

Here is how I would choose between the formats.

What the completion numbers actually show

Cohort courses are often said to reach completion rates above 90%, against about 3% for self-paced courses. Neither number holds up as a comparison.

The 90% figure traces back to a statement by the founder of a company that sells cohort courses, with no study, no sample and no definition of completion. The 3% comes from real research, but it is not about self-paced courses in general. In a 2019 analysis in Science, Justin Reich and José Ruipérez-Valiente looked at every MIT and Harvard course on edX up to May 2018. In 2017 to 2018, 3.13% of participants (30,650 of 980,589 people who registered and viewed a course) completed it. Among learners who paid for a verified certificate, 46.02% completed (21,321 of 46,326).

Those courses were free and open to anyone. Half of everyone who registered (52%) never opened the course material, and those people are not even in the 3% figure. Among participants who said they intended to finish, 15.42% did (6,478 of 42,005). The gap compares differently selected learners within these courses. It cannot isolate a pacing effect or establish why each person completed. Comparing a free open course with a paid, selective cohort programme measures the learners as much as the format. Medians across many open online courses are higher than 3% but still low: Katy Jordan's 2015 analysis found a median completion rate of 12.6% across 129 courses.

The better evidence on pacing compares two versions of one course. The University of Helsinki ran a free programming course both with a fixed schedule and as a self-paced course.

In Ihantola and colleagues' 2020 study, about 17% of the fixed-schedule learners dropped out in the first week, against about 50% in the self-paced version. After that, the authors write, "exit behavior is almost similar". Most of the gap opened in week 1. The MIT thesis found a smaller gap: when one MITx course moved from weekly releases and due dates to self-paced, with nothing else changed, completion among paying learners fell by about 5 to 7 percentage points.

A shared schedule is associated with more people staying active, especially at the start. That is not proof that schedules cause completion. And a schedule the learner sets for themselves is a different thing: in a randomised trial with 18,043 learners, Baker, Evans and Dee asked people to plan when they would study, and those who were asked were nearly 10% less likely to earn a certificate, a relative drop from a base of 9.3% in the control group. The schedules linked to more people staying active were set by the course and applied to everyone.

Cohort-based learning examples

Cohorts are common in education and in workplaces, often under other names.

• University and graduate school cohorts: students who enter a degree, a master's or a doctoral programme together and move through it as a group.
• Graduate intakes: a firm's new graduates who start on the same day and share a structured first year or first few months.
• New-hire onboarding classes: a group of new starters, such as a contact centre intake, trained together before they join their teams.
• Leadership programmes: a group of managers who meet over several months, with work between sessions.
• Career-change and job training academies: a full-time programme that trains a group and places them with employers, such as Year Up, a US job training programme.
• Teacher and clinician professional learning: small groups of colleagues who meet on a cycle to observe and discuss each other's practice, such as Quality Teaching Rounds in NSW public schools.
• Paid online courses with an intake: a course that runs from a set start date, with weekly modules, a group space and live calls.

Onboarding intakes are cohorts by another name

Organisational researchers studied cohorts long before online courses. In their 1977 working paper on how organisations socialise newcomers, John Van Maanen and Edgar Schein called it collective socialisation: "taking a group of recruits who are facing a given boundary passage and putting them through a common set of experiences together." They noted that recruits in a group "learn more from one another than from the faculty." They also gave the trade-off: collective tactics are "most likely to produce a custodial" orientation, in which newcomers learn to keep doing things the way they have always been done.

Later evidence suggests the group is not the strongest lever. A 2007 meta-analysis of socialisation tactics by Saks, Uggerslev and Fassina found that "the social tactics (serial and investiture) were the strongest predictors" of newcomer adjustment. Serial tactics give newcomers experienced role models; investiture affirms what newcomers already bring. So an intake is likely to do more when it also connects graduates with experienced colleagues and values what they arrive with, rather than relying on the group alone. That fits situated learning theory, in which newcomers learn by taking part in the work of an experienced community.

Year Up is not onboarding inside one employer, but it has the longest measured follow-up of any cohort programme I found: a US job training programme built around incoming cohorts.

Case study: Year Up job training cohorts in nine US cities

Year Up is a national nonprofit that trains young adults from low-income backgrounds for entry-level jobs in IT, finance and business operations, then places them in internships with employers. A federal evaluation followed its core programme in offices across nine cities for seven years.

• What they did: Each incoming cohort was organised into learning communities of about 40 participants and staff "to foster supportive social connections". Participants spent six months full time in training on technical, business communication and professional skills, then six months in an internship four and a half days a week, with a weekly half-day back with the group to share what was happening at work and plan their careers. Advisers, mentors and stipends were part of the package.
• Evidence and result: In the randomised trial reported by Fein and Dastrup for the US Administration for Children and Families in 2022, 75% of the treatment group completed the programme. In quarters 23 and 24 after random assignment, about six years on, average quarterly earnings were US$1,895 higher than the control group's US$6,901, a 28% increase, using wage records for 1,637 treatment and 858 control members. The evaluators estimated a return of US$2.46 per dollar spent.
• Limitation: The cohort was one part of a bundle with internships, stipends and advising, so its own contribution cannot be separated, and it is a pathway into employers rather than onboarding inside one firm. Year Up accepted close to one in six applicants, and the programme cost about US$28,290 per participant.
• Lesson: For a long, high-stakes intake, make the group the backbone and keep it meeting after people start real work, as Year Up's weekly half-day back did. Credit any measured gain to the whole programme, not to the cohort alone.

The five elements of a cohort, worked through one graduate intake

From here on I'll use one hypothetical example. An engineering consultancy that designs water and transport infrastructure for councils and state agencies has offices in Brisbane and Perth, two hours apart all year because neither city uses daylight saving. Its February intake is 16 graduate engineers, civil, structural and environmental, with 10 in Brisbane and 6 in Perth.

The graduates hold engineering degrees, but they do not yet know the firm's quality system, how a design is checked and verified before it goes to a client, how to book their time to project codes, or how to write for a client. A degree does not mean they can write a client memo, so the programme teaches it. The intended outcome is that by week 10 each graduate can follow the firm's check and verify process, write a client memo that passes the firm's checklist, book time correctly and know who to ask in both offices. The programme does not teach engineering theory or certify technical competence.

The firm could put all of this in a self-paced course. It chooses a cohort for four reasons: all 16 start on the same day; it wants a network across offices and disciplines; it wants graduates to review each other's client writing; and the programme lead wants to see the whole intake's work at the same points. The programme runs for 10 weeks and takes about 3 hours a week. Each element below gets the same three parts: what goes into the graduate intake, how to check it, and a common slip.

1. A shared start

The shared start is what makes a group an intake. Everyone begins on the same date, meets the same people in week 1 and moves through the weeks in step. It also creates the format's main problem for anyone who arrives late.

• In the graduate intake: all 16 graduates start on the same Monday in February. Week 1 is a welcome, a first live session and an introduction post to their small group.
• How to check it: list who starts on day one and decide now what happens to anyone who joins later. A graduate who starts in March either waits for the next intake or works through the material self-paced and joins a group for the remaining tasks.
• Common slip: calling a rolling programme a cohort. If people join every week, they do not share a start, and the group tasks fall apart.

2. Shared pace and deadlines

Shared pace means the weeks open in order and work is due at the same time for everyone. In the Helsinki comparison, most of the difference between the fixed-schedule and self-paced versions appeared in the first week, so the first weeks deserve the most care. Releasing content over time is often called drip-feeding, and a cohort adds the group deadline on top.

• In the graduate intake, new material opens each Monday for about 2 hours of self-study. Allow another hour for peer work in non-live weeks, or 45 minutes live and 15 minutes peer work in live weeks. This is a proposed 3-hour budget to test. Each week has one group task with a due day, such as replies due by Thursday in week 2.
• How to check it: for each week, write down what opens, what is due and on which day. If a week has nothing due, ask why the group needs to be together that week.
• Common slip: deadlines so rigid that busy people give up. A graduate on a site visit for three days needs a way to catch up without losing their place, such as a two-day grace period announced in week 1.

3. Peer work in small groups

A shared cohort rhythm can make planned peer work easier to organise, although self-paced courses can also include peer interaction. It has to be designed. A 2012 review by Borokhovski and colleagues separated distance courses that only made interaction possible, which "contain the necessary conditions for student-student interaction to occur, but are not intentionally designed to create collaborative learning environments", from courses with designed interaction tasks. "The results favored designed interaction treatments over contextual interaction treatments." A forum on its own is not peer work. A task that needs other people is.

Small fixed groups, often called pods, make that work manageable. Alfred Rovai, in a 2002 essay on building a sense of community at a distance, suggested small groups "typically under ten learners each", with specific tasks and timelines. Treat that as a practitioner's rule of thumb, not a finding.

• In the graduate intake: four pods of four, mixed across the two offices and the three disciplines, the same pods for all 10 weeks. In week 2 each graduate posts a calculation, drawing or report section they were asked to check or produce that week: what they checked, what they had to ask about and who they asked. They then reply to two pod members, at least one from the other office, by Thursday. In week 6 each graduate writes a two-page technical memo to a council client and reviews the memos of two pod members against the firm's five-item memo checklist: purpose stated, assumptions listed, figures traceable, recommendation clear and plain language.
• How to check it: for each pod task, ask whether a graduate could complete it alone. If they could, it is individual work with a forum attached.
• Common slip: grading discussion by counting posts. Two posts a week is easy to count and says nothing about whether anyone learned from them.

Our guide to the Community of Inquiry framework covers how to design and facilitate online discussion in more depth, and our guide to collaborative learning covers group tasks more broadly. A randomised Australian trial shows what a small fixed group can do when it has a shared standard and a strict routine.

Case study: Quality Teaching Rounds in NSW public schools

Schools spend heavily on teacher development, with little evidence that it changes teaching. Researchers at the University of Newcastle, working for the NSW Department of Education, tested Quality Teaching Rounds, in which small fixed groups of teachers observe and discuss each other's lessons.

• What they did: After a two-day workshop, teachers formed professional learning communities of 3 to 8 who stayed together for a full set of Rounds over two school terms in 2014. Each Round took a day: a discussion of a reading, the whole group observing one member's lesson, each teacher coding the lesson against a shared Quality Teaching framework, then a group discussion. No outside expert ran the groups; the framework and a fixed protocol did that work.
• Evidence and result: In a cluster randomised trial across 24 schools and 192 teachers, observers who did not know which group a teacher was in scored 1,073 lessons. Gore and colleagues' 2016 report found a positive effect on teaching quality at six months (d = 0.4, an effect size usually read as moderate), still present at twelve months (d = 0.2 and 0.5 in the two versions tested), and "Higher fidelity (set at 6 of 8 criteria) produced stronger effects on quality of teaching". There was no significant effect on trust or in rural schools.
• Limitation: The programme's developers ran the evaluation, the schools volunteered, and the outcome was observed teaching, not student results. The effect belongs to the whole package of framework, readings, observation and fixed group, and the source calls these groups professional learning communities rather than cohorts.
• Lesson: A small fixed group changed observed practice when it had a shared rubric and a routine strict enough to run without an expert, and groups that followed more of the routine did better. Give each pod the checklist and the routine, then check they follow it.

4. Live touchpoints

Live touchpoints are the times the group meets at once, in a room or on a call. They are optional in a cohort, but they do a job the asynchronous parts cannot: questions answered on the spot, practice in front of colleagues, and a chance to watch peers present and be questioned. Learning by watching how others handle a task, and what happens to them, is central to social cognitive theory.

• In the graduate intake: a 45-minute live video session every second Thursday at 12:30 pm Brisbane time, which is 10:30 am in Perth, in weeks 1, 3, 5, 7 and 9. Each session is recorded for anyone on a site visit. In week 7 the session is a design review run the way the firm runs one: two graduates present their memo, the principal engineer asks the questions a checker would, and the pods discuss what changed.
• How to check it: pick the time from the learners' calendars and time zones, not the facilitator's. Then decide what a graduate who misses it must do, such as watching the recording and posting one question to their pod.
• Common slip: filling the live time with a presentation the learners could have read. Use it for what needs people at once: questions, practice and review.

5. The facilitator

The facilitator runs the cohort: opens each week, reads the groups' work, replies to quiet groups and chases late work. This is the element with the clearest cost. In the maths teachers trial, facilitators spent about 11 hours a week on each cohort class and about 3.5 on each self-paced class.

• In the graduate intake: the graduate programme lead, a principal engineer, with about 3 hours a week set aside. They run the live sessions, read every pod's thread and review the peer reviews rather than every memo. Each graduate's line manager signs off the week 8 task, a first piece of real project work checked through the firm's process. Three hours is tight against the 11 in the maths trial; the proposed budget relies on the lead reviewing the reviews rather than every memo, and must be tested and the pods carry most of the replies, so budget more in weeks 6 and 7.
• How to check it: add up the facilitator's hours for each week and put them in their diary before the intake starts. If no one has the time, plan a self-paced course with a few group tasks instead.
• Common slip: naming a facilitator who has no time. The group then waits for replies that do not come, and the pods stop posting.

How much of each element do you need?

Not every cohort needs all five elements in full, and they do not have to be added in a set order. They affect each other, so the right mix depends on the learners and the job.

• Live time depends on schedules. For two offices two hours apart, a session at 12:30 pm in Brisbane, 10:30 am in Perth, suits both. For shift workers across three time zones, an asynchronous cohort with a recorded call may be the only option.
• Group size depends on dropout. Pods of four become pods of two if two people leave. Start with four or five, and plan to merge groups that shrink. Rovai suggested that eight to ten learners is about the minimum for good discussion in a whole online class, and 20 to 30 about the most one instructor can handle in active discussion. Those are his estimates, not findings.
• Facilitator time depends on the peer work. If the facilitator reviews the reviews rather than every piece of work, one person can support more pods.
• The need for a cohort depends on the task. If nothing in the programme needs other people, a self-paced course with a few check-ins may be simpler; compare the actual delivery costs.

The sense of belonging that a cohort can create also has a place in motivation research. Self-determination theory treats relatedness, the need to feel connected to others, as one of three basic psychological needs. A cohort can meet it, but only if people actually interact.

How to run a cohort in workplace training

This workflow is mine. It turns the five elements into a plan, and it follows the graduate intake.

1. Check that the job needs a cohort. Ask whether learners start together, whether some of the work needs other people, whether someone has facilitator time each week and whether the group can meet or keep in touch on a schedule. The graduate intake passes on all four. A compliance refresher for staff hired all year round would not, so it stays self-paced.
2. Fix the intake and the pods. Set the start date, the size and the groups. Mix pods across offices or teams on purpose, and decide what happens to late starters and to a pod that drops to two.
3. Write the peer tasks and their due days. In most weeks, one task that needs the pod, such as the week 2 post and replies and the week 6 memo review. Give each task a standard to judge against, such as the five-item memo checklist.
4. Build the content around the tasks. Release it on a weekly rhythm, and keep each week to the time you promised, about 2 hours of self-paced work in the graduate intake.
5. Plan the live touchpoints. Set times both offices can make, a recording, and a job for each session that needs people at once.
6. Give the facilitator a weekly routine. Open the week, read each pod's thread, reply to quiet pods, check the peer reviews and chase late work.
7. Bring line managers in. Tell them the schedule, and give them one task to sign off, such as the week 8 piece of checked project work.
8. Decide how you will judge it. Look at the work graduates produce, such as memos that pass the checklist and checked project work, rather than a satisfaction survey alone. Our guide to the Kirkpatrick model sets out how to evaluate a programme beyond reactions.

The week-by-week plan for the graduate intake

Here is the 10-week plan. Each week opens on a Monday; live sessions are every second Thursday at 12:30 pm Brisbane time, 10:30 am in Perth.

The week 6 memo uses illustrative figures supplied in the course: a council asks for a preliminary cost and fauna-passage comparison of two culvert options for a road upgrade, a 1,200 mm pipe culvert estimated at $180,000 or a 2,400 mm by 1,200 mm box culvert estimated at $260,000 that also allows the fauna passage the council has asked about. Graduates use those facts for a writing exercise and state that hydraulic capacity, approvals, site conditions and engineering suitability have not been assessed. The memo cannot recommend a final design from dimensions and cost alone.

What Coursebox course requests leave out

To see whether people think about delivery when they ask for a course, I looked at course requests typed into Coursebox in the 90 days to 7 October 2026: one per person, from English-language browsers, with no uploaded file. Of 266 prompts, 104 were course requests. Two independent AI readers each classified all 104 against written definitions of the cohort elements, and I re-read every positive call in full. Counts combine requests flagged by either reader. They are classification estimates, not statistical upper bounds.

People describe what a course should cover and leave out how people will take it. The word "cohort" appeared in none of the 104 requests, and "self-paced" in one, for a teens' internet safety course. Thirty-one requests were for workplace or professional training (27 if you count only those both readers agreed on).

Even a schedule is not a cohort. Eighteen requests laid the course out in days or weeks, a classification estimate, and half of them mentioned no group, live session, peer task or trainer at all. The only request that described the pieces of an online cohort together, a paced eight-week course with regular live Q&A sessions and a private forum, was for a hobby course.

The sample is small, and it covers what people typed, not what their organisations do around the course. Some may run intakes, live sessions and managers' check-ins that never appear in a prompt. Still, a builder given only the content, whether a person or an AI, is likely to produce a self-paced course, and "Week 1, Week 2" headings do not make a cohort. If you want a cohort, write the start, the pods, the peer tasks, the live sessions and the facilitator into the brief yourself.

Disadvantages of cohort-based learning

Cohorts cost more to run than self-paced courses, and they have other drawbacks.

• More facilitator time. In the maths teachers trial, each cohort class took about three times the facilitator hours of a self-paced class.
• More dropout is possible. In the same trial, 26 teachers left the cohort and 18 the self-paced version. A fixed pace does not suit everyone's workload.
• Rigidity and personal cost. A 2000 study of cohorts in educational leadership programmes by Barnett and colleagues put the two sides plainly: "Cohort users point to organizational efficiency and learner benefits; nonusers deplore rigidity and personal costs for students."
• Late starters wait. Someone hired a week after the intake starts either waits for the next one or joins without the shared start.
• Scheduling across offices and shifts. Every live session excludes someone, so recordings and asynchronous alternatives are part of the design, not an afterthought.
• Deadlines can backfire. A 2019 study by Miller, Asarta and Schmidt found that rigid deadlines "detracted from student participation". I read only the abstract. In the Helsinki comparison, the schedule's advantage appeared almost entirely in the first week, so deadlines may help most at the start and do little after that.
• Groups can keep the old way. Van Maanen and Schein warned that collective socialisation tends to produce newcomers who keep things as they are. Pair the intake with experienced role models and invite graduates to question how things are done.
• Pods depend on their members. A pod with one quiet member, or one who leaves, does less for everyone in it. In my view, the facilitator's most useful weekly job is spotting these pods early.

Building a cohort course in Coursebox

Coursebox also has an AI LMS for enrolling learners and running courses, but for this post I tested only the course side: whether the AI course builder keeps a cohort's group work, or flattens it into a self-paced course.

On 12 October 2026 I briefed Coursebox to build the graduate intake course. Coursebox is our product, so weigh my account with that in mind. The brief described the 16 graduates and the two offices, the Monday release, the live sessions in weeks 1, 3, 5, 7 and 9, the pods of four and the facilitator, gave the culvert figures and the five-item memo checklist, and asked for 11 sections in order, from a short "Before you start" to a wrap-up, with a separate section of facilitator notes. It asked for open-text posts and assignments rather than multiple-choice quizzes, and said not to claim that weeks open automatically on dates.

The first outline was far bigger than the brief: 22 sections, 101 pages and 9 quizzes, with weeks split across several sections. I sent one revision asking for exactly 11 sections, no quizzes, a discussion in week 2 and the peer review in week 6. The builder asked once which structure to use, and I chose the 11 named sections. The second outline had the 11 sections in the order I asked for, 21 pages and no quizzes.

Most of it then worked. The page on how the 10 weeks run had a weekly rhythm table and a "What is due when" list, and it made no promise that weeks would open on dates. The week 6 page was an assignment asking each graduate to write the culvert memo and review two pod members' memos against the five-item checklist. The facilitator notes came back as their own section, and the wrap-up was last.

The issue was week 2. Its two pages, "Post what you checked" and "Reply across offices", explained how to post and what to ask, with a pattern to follow, but there was nowhere to post or reply. The whole course had one assignment, in week 6, and no discussion at all. The shared task that makes the pods a cohort had become instructions for a conversation the course could not hold.

I checked the outline, the rendered pages and the settings of one block. I did not set up groups, test learners, calendar events or a peer review run, so I am not describing those features here, and I did not run the intake with graduates. The block I added is one discussion for everyone enrolled, not one per pod; I did not test splitting it by pod. A discussion box is not a discussion: the graduates still have to post and reply, and the facilitator still has to read it.

To fix week 2, I opened the "Post what you checked" page in the editor and added a Discussion block from the Interactions panel, titled with the week 2 posting task and its Thursday deadline. Its posting setting defaulted to admins and facilitators only, which would have let only the programme lead post, so I changed it to anyone can post, with enrolled users able to view. Then I reloaded the learner view: the page showed the discussion title and a "Share something" box with a Send button, ready for the first post.

Putting cohort-based learning to work

A cohort is a group that starts together and keeps to shared deadlines; live sessions are optional. The completion figures used to sell cohorts do not compare like with like, and the two trials that ran one course both ways found no significant difference between the versions, while in the one that counted facilitator hours, the cohort took about three times as many.

So choose a cohort for what only a group can give: peer review against a shared standard, discussion of real work across offices, a network for a new intake, and a programme lead who can see everyone's work at the same points. Keep self-paced courses for reference knowledge, refreshers and people who start at different times.

Your next step: take the programme you are planning, write down one task for each week that a learner could not do alone, and give it a due day. If you cannot find those tasks, run it self-paced.