Why a pod needs a range in it
Four students who all understand the material have nothing to ask each other. Four who are all lost have nobody to ask. So the class is split into quarters by how well students understand the material, and every pod takes one student from each quarter — that is what guarantees the mix. The seating is deliberate rather than random. Shoulder partners are two quarters apart (Group 1 with Group 3, Group 2 with Group 4) and face partners are one apart (Group 1 with Group 2, Group 3 with Group 4). Group 1 and Group 4 never work together directly — putting them together sounds like the helpful thing to do, but that gap is usually too wide: the explanation lands over the other student's head and it becomes one student working while the other copies. Capping every pairing at two quarters keeps all of it close enough for an explanation to get through.
If you just want groups, stop here. Generate the pods, read them off the board, and you are done — the rest of this is for teachers who want to seat the pods and use the partner pairs.
Pod of 4 — the default layout
↑ Far from teacher
Group 1
Already mastered it
↕ shoulder↕ shoulder
Group 4
Just starting on it
↓ Near teacher
Shoulder partners —
Group 1 with Group 3, and Group 2 with Group 4. Two quarters apart.
Face partners —
Group 1 with Group 2, and Group 3 with Group 4. One quarter apart: the closest pairing in the pod.
Nobody is ever paired with someone too far away. Group 1 never works directly with Group 4 — that pairing only exists on the diagonal. The widest pairing in the pod is two quarters and the narrowest is one, and that cap is the entire point: every conversation in the pod is close enough in understanding that an explanation actually works. The seating gets them near enough to explain; the structures do the rest.
Face partners give you two conversations at once — the top half of the room working together, and the back half working together. That makes it a check on your teaching as much as a grouping. Say
"Face partners, explain step three to each other." If your Group 1s and 2s cannot explain it to each other, the top half of the room has not got it and you need to reteach before you go on. At the same time Groups 3 and 4 are talking at their own level instead of sitting quiet while somebody well above them narrates.
Shoulder partners cross the halves — Group 1 with Group 3, Group 2 with Group 4. Two quarters apart, so this is the pairing where one student has genuinely more to unpack for the other. Use it when you want the explaining to travel further, and alternate it with face work so the same student is not doing all of it.
Then use a structure. Neither pairing produces talk on its own — proximity is not a task. Name the pair and give them a job with a required output:
"Face partners, explain step three, then I am asking one of you at random", or
"Shoulder partners, compare answers and write the one sentence you both agree on." One instruction moves the whole room and nobody negotiates who they are working with.
What to do once they are seated: the
Cooperative Learning Toolkit has the structures — a library of collaborative methods, an Activity Designer for building the task, and a Question Designer for the prompts you put in front of the pair.
Pod of 3 — when the numbers don't divide evenly
↑ Far from teacher
Group 1
Already mastered it
↓ Near teacher
When the class doesn't divide by four, whoever is left over forms pods of 3. One student sits farthest from you and the other two sit nearest, so a pod of 3 has one face pair on each side and a shoulder pair across the bottom. In practice, three students are close enough that everyone works with everyone — you can call a pair by name or just say "in your pod of three."
One thing to know: leftovers are taken from whichever groups still have students in them, so a pod of 3 will not always hold three different quarters the way a pod of 4 always does. Glance at the badges before you set a task on it, and drag names between groups if you want a different mix.
How to sort your class
Go by how much of this material each student has already mastered. Any evidence you already have works: their grade in your class, performance on tests and quizzes, what they earned on recent assessments, a pre-assessment for the unit, the last few assignments, or simply your read of where they are. The students who have it go in Group 1, the students just starting on it go in Group 4, and the rest split between Groups 2 and 3 — about a quarter of the class in each. It does not need to be precise; a reasonable read is the whole point. Drag anyone to a different group the moment your read changes.
What the research actually says
Worth being straight about this, because it changes how you use the tool. Putting students in small groups, on its own, does very little. Across 145 studies the average achievement effect of small-group work over whole-class teaching was about +0.17, and head-to-head comparisons of mixed-level against same-level grouping came out roughly even — in that same review, slightly in favour of same-level (+0.12). Hattie puts within-class grouping at d = 0.18.
What moves achievement is the cooperative structure laid on top of the grouping. Hattie reports d = 0.55 for cooperative versus individualistic learning. Slavin's review is sharper: team learning where the group's success depends on every individual member learning it has a median effect of +0.32, while informal group work with no such accountability sits at +0.07.
The mechanism is explanation. Students who give and receive elaborated explanations gain the most, and trading bare answers is negatively related to achievement gain. Mixing the quarters is what makes a real explanation possible — it does not cause one.
So the pods are the setup, not the intervention. Give each pod a goal it can only reach if every member learns the material, hold individuals accountable for their own understanding, and ask for reasoning rather than answers. One more finding worth planning around: Webb found that in mixed groups the higher-attaining students do most of the explaining — so the students who would gain most from explaining are the ones least likely to do it. Assign who explains rather than leaving it to the pod.
When to reshuffle
Reshuffle every 5–6 weeks so students work with different people across the semester. Your sort stays put — only the pod combinations change. And because the groups describe where someone is with the current material, expect the sort itself to shift between units: move a student to a different group before you regenerate, and the pods follow. The tool keeps your sort between visits.