You say the words every teacher has said. Get into groups of four. And in about nine seconds the room reorganizes itself along lines that have nothing to do with your lesson. The confident kids cluster. The best friends find each other across the room like magnets. And then there are the three or four students standing near their desks, doing that thing where they pretend to look for something in their bag so it does not look like they are waiting to be wanted.
Everyone in the room saw it. Including them.
That moment is the actual problem with grouping students, and it is worth being precise about what the problem is, because most advice gets it wrong. The problem is not that you need a random number. You can generate a random number in your head. The problem is that whoever decides the groups owns the outcome, and every method of deciding hands that ownership to someone who then gets blamed for it. Let the students choose and the social hierarchy of the room does the deciding. Choose yourself and you become the person who put Maya with the group she asked not to be in. Count off and the class quietly works out that you controlled where the counting started.
What follows is seven ways to group students, what each one is genuinely good at, what it costs you, and when it blows up. Then the practical part: how to run a random split in front of the class so that the sorting stops being your decision, and what to do about the student nobody picks.
Why Grouping Goes Wrong More Often Than It Should
Group work carries a weight that other classroom transitions do not. Being placed somewhere is a small public statement about where you sit in the room, and students read it that way even when no adult intends it. A teacher can move thirty kids into groups in under a minute and accidentally publish a seating chart of the class social order.
The three failure modes are worth naming because they explain why some methods keep disappointing you.
The visibility failure. Anything that happens where students can watch it gets judged. Student choice is maximally visible, which is why it hurts the most. But teacher assignment is visible too. You read the list out, and every student hears who you paired them with and starts building a theory about why.
The authorship failure. Someone made the call. If you made it, the complaints come to you, and they come in the form of negotiation. Can I switch. Can I just work with Sam. I cannot work with him. Once the door is open to negotiation you spend your lesson doing diplomacy instead of teaching.
The pattern failure. Methods that feel random often are not. Counting off around the room groups students by where they sit, and students sit near their friends, so counting off frequently produces the same clusters you were trying to break up. Alphabetical grouping produces the exact same groups every single time, all year.
The methods below get compared on those three axes, because a method that solves one and ignores the other two is what you have probably already tried.
The Seven Ways to Group Students, Honestly Compared
1. Let students choose
Fast, popular, and the students are happy with the result before the work starts, which is not nothing. Motivation genuinely rises when people work with someone they like.
The cost is the moment described at the top of this article. Student choice does not distribute students, it distributes social capital, and it does it publicly. It also tends to group by ability as a side effect, because friendship groups in a classroom often form along lines of confidence with the subject.
Worth using for: short, low stakes activities where the pairing barely matters and speed does. A two minute think and share. A quick practice round.
Avoid for: anything graded, anything long, and any class where you already know some students will be left standing.
2. Assign the groups yourself
The strongest method for actually engineering an outcome. You know who needs support, who works well together, who cannot be within two metres of each other before someone cries. Nothing else gives you that control.
The cost is that you are now the author. Every objection arrives at your desk, and you have to defend decisions you made for reasons you may not want to say out loud in front of the class. You cannot explain that you separated two students because one of them dominates the other. So you end up saying because I said so, which is true but corrosive if it is your only answer all year.
It also takes real time. Twenty eight students into seven balanced groups, done properly on paper, is a fifteen minute job you do at home.
Worth using for: long projects, groups that will exist for weeks, situations with genuine pastoral considerations, and any class where you know a random result could go badly for a specific child.
Avoid for: everyday grouping, because you will not sustain it and the negotiation tax gets exhausting.
3. Count off around the room
The oldest trick in teaching. Everyone counts one two three four, and all the ones go together. It costs nothing and takes forty seconds.
The problem is that it is not random. It is a function of the seating plan. Students sit near people they chose to sit near, so counting off tends to slice friendship clusters into groups in a fairly predictable way, and clever students work out very quickly that they can shuffle two seats to land on the number they want. Watch a class during a count off and you will see it happen.
Worth using for: quick transitions where approximate mixing is fine.
Avoid for: any time the result matters, or with a class that has already gamed it.
4. Draw the groups at random, in front of everyone
The class list goes into a picker, you set the number of groups, and the sorting happens live on the board while everyone watches. Nobody chose. Not the students, not you.
The reason this works is not the randomness, it is the witnessing. When the sorting happens in front of the room, the outcome has no author to argue with. Complaints stop being negotiations and become comments about luck, which is a completely different conversation and a much shorter one. It also removes the quiet cruelty of student choice entirely, because nobody is ever picked last when nobody is picking.
The cost is that random results are sometimes bad results. You can end up with a group of four who will produce nothing, or a student placed with someone they have a genuine reason to avoid. You need a plan for that, and there is one further down.
Worth using for: most everyday grouping, especially with classes where the social dynamics are the main obstacle.
Avoid for: high stakes long projects without at least a quiet review of the result first.
5. Group by ability
Sorting students by current attainment so you can pitch the task at each group. It makes differentiation practical, and for some tasks it genuinely serves students better than mixed groups do.
It is also the most contested method on this list, and it deserves an honest note rather than a confident one. Long running attainment grouping has a well documented tendency to become self fulfilling: students learn which group they are in, and so does everyone else, and expectations follow the label. The evidence on it is genuinely mixed and depends heavily on whether groups stay fixed and how visible the sorting is.
If you use it, the practical advice most teachers converge on is to keep it task specific and temporary rather than a permanent structure, and to avoid naming the groups in any way that ranks them.
Worth using for: a specific skill practice session where the task really does need different entry points.
Avoid for: a default structure that persists across a term.
6. Group by interest or by choice of topic
Put the topics on the board, let students pick the topic, and the groups form as a side effect. The grouping is genuinely incidental, and students bring real motivation because they chose the subject rather than the people.
The cost is that you get no control over group size, and popular topics create groups of nine while somebody sits alone with the topic nobody wanted. It also quietly reintroduces friendship grouping, because students watch where their friends go and follow.
Worth using for: research projects, presentations, anything where genuine interest carries the work.
Avoid for: when you need even group sizes, which is most of the time.
7. Fixed groups that rotate on a schedule
Set groups at the start of a unit, keep them for the whole unit, then reshuffle. Students learn to work with their group properly rather than restarting the social negotiation every lesson, and they build the actual skill of collaborating with people they did not choose.
The cost is commitment. A group that is not working is not working for six weeks, and you will have to intervene rather than wait it out. It also needs a genuinely good initial sort, which usually means method two or method four.
Worth using for: units with sustained group tasks, and for building collaboration as a skill in its own right.
Avoid for: classes with volatile dynamics where six weeks is a long time.
The short version. Use random grouping as your everyday default because it removes you as the author. Use deliberate assignment when the stakes are high enough to justify owning the decision. Use student choice only when the pairing genuinely does not matter. Everything else is a variation on those three.
How to Run a Random Split In Front of the Class
The mechanics matter more than they look, because the entire benefit comes from the class seeing it happen. A random split done privately at your desk gets you none of the trust and all of the risk.
The setup, once, at the start of the year:
Get your class list in once and save it. Type or paste the names, one per line. Most teachers paste straight from a gradebook or a spreadsheet column. Save it to an account and it follows you from the laptop you plan on at home to the machine bolted to the classroom wall, so you never type the roster twice. If you teach several classes, make one wheel per period and name them plainly.
Put it on the board. Cast, mirror or just open it on the projector. It has to be visible. This is the entire mechanism.
Then, every time you group:
Pick your number of groups, not your group size. Teams mode asks how many groups you want, then distributes everyone evenly across them. If you have twenty eight students and you want groups of four, that is seven groups.
Say what is about to happen before it happens. One sentence. We are drawing groups, the wheel decides, and we are not renegotiating after. Saying it in advance is what closes the door on the negotiation, and it is much easier to say before the result exists than after somebody does not like it.
Spin once, and let it play. The sorting animates and the room watches names land in groups. Students react, and the reaction is the point. It is loud for about eight seconds and then it is settled, because there is nobody to appeal to.
Leave the result on screen while they move. It stops the six students who ask which group am I in again, and it means the record of what was drawn is public rather than something you are remembering.
You can run all of this from the random team generator, which is built for exactly this and free with no account needed to try it. If you want it framed for a classroom specifically, with the class list saving and the projector setup, the classroom wheel is the same engine set up for a room full of students.
One sentence that saves you the whole argument. The wheel decided, and I am not overruling the wheel. It sounds almost silly written down. In a classroom it works, because it moves you from being the person who made the decision to being the person who is also subject to it.
What to Do About the Student Nobody Picks
Every teacher has at least one. Sometimes it is the child who is genuinely struggling socially. Sometimes it is a new student who arrived in March. Sometimes it is a kid who is fine in every other respect and simply has not landed with this particular group of thirty.
Student choice grouping is brutal for that child, and it is brutal in a specific way: it is not a single event, it is a repeated public confirmation. Once a term is survivable. Every Tuesday is a different thing entirely.
This is the strongest argument for random grouping, and it is worth stating plainly. When the wheel sorts the room, nobody is chosen last, because nobody is choosing. The child who would have been left standing simply appears in a group like everyone else, and the moment passes without anybody noticing it, which is exactly what that child needs. Not a kind intervention. Not being assigned a buddy. Just an ordinary group, arrived at the same way everyone else got theirs.
Teachers who switch to random grouping often report this as the change they did not expect and would not go back on. The efficiency is fine. The removal of that recurring moment is what actually matters.
Two practical additions. If a student has a specific reason they cannot work with someone, deal with it before the draw rather than after, quietly, by adjusting who is in the pool for that session. And if a random result lands somewhere genuinely unworkable, you are still the teacher. Say the wheel gave us something I need to adjust, make the one change, and move on. Doing that occasionally does not break the system. Doing it every time does.
Group Size, and Why Four Is Usually Right
Group size changes behaviour more than most teachers expect, and the differences are consistent enough to plan around.
Pairs. Maximum participation, because nobody can hide in a pair. Ideal for practice, drilling, checking work, and quick discussion. Weak for anything requiring division of labour, and one absence destroys the group entirely.
Threes. Deceptively difficult. Threes tend to produce a pair and a spare, and the dynamic can be unkind in a way that is hard to see from the front of the room. Useful when the task genuinely has three distinct roles, and awkward otherwise.
Fours. The default for good reason. Large enough to split work meaningfully, small enough that nobody can disappear, and it breaks cleanly into two pairs when part of the task suits that. If you are unsure, use four.
Fives and above. Passenger territory. In a group of five there is almost always one student who contributes nothing and is difficult to catch doing so, because four people are visibly working. Use larger groups only when the task genuinely has five or more distinct jobs and you assign them explicitly.
With an awkward class size, uneven groups are better than forcing it. Twenty six students into groups of four gives you six groups of four and one group of two, and a pair is a perfectly good group. Trying to make it come out even is how you end up with sevens.
When Random Grouping Is the Wrong Call
Recommending one method for everything is how advice stops being useful. Random grouping is the right default, and here is where it is not.
Assessed group work with a real grade attached. When the group's output determines an individual's mark, the fairness of the grouping matters far more than the fairness of the draw. A random result that puts four struggling students together and four confident ones together produces two very different grades from the same task. Sort that one deliberately.
Students with support plans or specific needs. If a student needs a particular partner, a particular seat, or a particular adult nearby, that requirement outranks the elegance of a random draw. Handle it before the draw, not by overruling it afterwards.
Groups that must survive a long project. Six weeks is long enough for a bad combination to do real damage. Either sort deliberately, or draw randomly and then review the result privately before announcing it, which keeps most of the benefit.
The first week with a new class. Before you know anything about the room, a random draw can walk you straight into a situation you had no way to anticipate. Early on, deliberate grouping while you learn who is who is the more careful choice.
Immediately after a conflict. If two students had a serious incident this morning, today is not the day to leave their pairing to chance in front of an audience.
Making It Routine
The methods only pay off if they survive a busy Tuesday. Three things make grouping stick as a habit rather than a project.
Set it up once. Rosters saved per class, one wheel per period, named so you can find the right one in two seconds while thirty people watch you. If you have to retype a class list you will stop doing this by week three.
Use the same words every time. Students settle into a routine far faster when the language is identical. We are drawing groups. The wheel decides. Find your table. Predictability is doing a lot of the work here.
Do not explain it twice. The first time you use a random draw, explain briefly why. After that, just do it. Continuing to justify it signals that it is still up for debate.
One extra idea worth stealing: use the same draw for jobs inside the group. Draw the groups, then draw the roles. Whoever is picked first is the recorder, second is the presenter, and so on. It solves the second negotiation of the lesson using the machinery already on screen, and it takes ten seconds.
The Point of All of This
Grouping students is not really a logistics problem, which is why logistics solutions keep leaving something unresolved. It is a question of who carries responsibility for an outcome that everyone in the room can see and has feelings about.
Take that responsibility off both the students and yourself, put it somewhere visible and obviously indifferent, and most of the friction disappears. Not because the groups are better. Because there is nobody left to argue with.
And a child who used to spend that thirty seconds looking through their bag gets to just be in a group, like everybody else.
Draw your groups in front of the class
Paste your class list once, pick how many groups you need, and let the room watch the sorting happen. Free, no account needed to try it.
Open the team generatorFrequently Asked Questions
Random grouping is fair in the sense that matters most in a classroom, which is that no person decided the outcome. It does not guarantee balanced groups, and it can occasionally produce a combination that will not work well. The fairness is procedural rather than in the result: nobody was chosen last, nobody was favoured, and there is no author to appeal to. For everyday grouping that is usually the fairness students care about. For assessed work where the group determines a grade, sort deliberately instead.
Use a method that does not depend on where students sit or what order their names are in. Counting off groups students by seating, and students sit near their friends, so it recreates the same clusters. Alphabetical grouping produces identical groups all year. A genuine random draw from the full class list is the reliable way to break the pattern, and because it is re-drawn each time, the combinations keep changing.
Four is the safe default. It is large enough to divide work meaningfully and small enough that nobody can hide. Pairs give maximum participation and suit practice tasks. Threes often become a pair and a spare. Groups of five or more almost always carry a passenger. With an awkward class size, one uneven group is better than forcing everything to come out even.
Stop using student choice for anything that matters. Letting students pick is the method that publicly confirms who is wanted and who is not, and for the child who is never picked it repeats every time. A visible random draw removes that moment completely, because nobody is picked last when nobody is picking. The student who would have been left standing simply appears in a group like everyone else.
Yes, and you should. The entire benefit comes from the class watching the sorting happen, so it needs to be on the board rather than on your laptop screen. It runs in any browser with nothing to install, so a projector, a Chromebook cast to the display, or an interactive whiteboard all work the same way.
Draw the groups once at the start of the unit, then screenshot or save the result and keep using it. Fixed groups that last several weeks let students actually learn to work together rather than restarting the social negotiation every lesson. The trade off is commitment: a group that is not working is not working for the whole unit, so review the draw before you commit to it.
Handle it before the draw, not after. Quietly leave one of them out of the pool for that session and place them yourself, or adjust the list so the pairing cannot happen. Overruling the result after everyone has seen it undoes the thing that makes the method work, because it tells the class the draw is negotiable. Doing it occasionally is fine. Doing it every time is not.
Let the groups be uneven. Twenty six students in groups of four gives you six groups of four and one pair, and a pair is a perfectly good group. Forcing the numbers to come out even is how you end up with groups of seven, which is worse than one small group. A team generator distributes everyone across the number of groups you ask for and handles the remainder itself.