Grouping Students: By Number of Groups or By Size

Group by the number of groups when the room sets the constraint — five tables, five groups. Group by size when the task sets it — lab kit for three, groups of three. The two modes sound interchangeable and are not: with 26 people they produce different groupings, and the remainder lands differently in each.

By number of groups

You say "four groups", and the class divides into four, whatever size that makes them. 26 people become 7, 7, 6, 6. This is the mode for fixed stations, tables, or debate sides — the structure is what it is, and the group size flexes to fill it.

By people per group

You say "groups of four", and the list divides into as many full groups as it supports. 26 people become six groups of four and one of two — unless the tool is sensible about it and folds the leftover pair into existing groups, making 4, 4, 4, 4, 5, 5. The task defines the size — pair work, lab benches, peer review triads — and the number of groups falls out.

The remainder is the real difference

By-count spreads the remainder automatically, so sizes never differ by more than one. By-size has to decide what to do with the people left over, and a tool that strands two people as their own "group" has made your decision for you, badly. Watch for that; it is the commonest failure in grouping tools.

Both modes, one paste

The Random Group Generator offers both — paste the roster once, pick the mode that matches the constraint you actually have, and reshuffle until the combination works. The underlying fairness is the same either way: shuffle, then deal.