Random vs. ability grouping: the honest trade-offs

Few staffroom debates run longer than this one, and both camps argue as if the other has no case. The honest answer is less comfortable: neither method wins outright, both have real costs, and the choice should belong to the task in front of you, not to a philosophy adopted in September.

What the evidence actually supports

Research on within-class attainment grouping points to modest effects on results at best, and a consistent warning about the side effects. Grouping by ability tends to help higher attainers a little and does least for the students placed lowest, whose confidence, expectations, and diet of work all quietly shrink. The academic label has a way of becoming a personal one.

It's also worth separating two things the debate usually blurs: setting between classes and grouping within one. The evidence concerns are strongest for the fixed, long-lived versions of both. Within a single lesson, where you control the task and the timescale, the risks are smaller and the flexibility is greater, which is exactly why the within-lesson choice deserves an actual decision rather than a habit.

Random grouping doesn't directly boost attainment either; that isn't its job. What it reliably delivers is breadth (everyone works with everyone), no status hierarchy baked into the seating, and expectations that stay flat across groups. Its cost: on any given day, chance can hand you a group that lacks the one skill the task needed.

Where ability grouping earns its place

Grouping by current attainment works best when it is specific, short, and temporary: twenty minutes on quadratic factorising with the five students who bombed yesterday's exit ticket, while everyone else moves on. The grouping is diagnostic, it dissolves when the gap closes, and next lesson the groups are different.

Where it goes wrong is permanence. The "circles table" set in autumn is still the circles table in summer, everyone in the room knows exactly what the tables mean, and students describe themselves by their table. At that point the grouping isn't responding to attainment; it's manufacturing it.

Where random is the better default

• Discussion and oracy: mixed voices beat matched ones, and hearing different thinking is the point.
• Projects and practicals: teams in the world outside school aren't sorted by test score, and mixed groups spread skills like organisation and calm that tests don't measure.
• Classroom culture: a visible, fair draw kills the "good group / bad group" reading before it starts, and re-dealing regularly means no combination hardens into an identity.

Four questions before you group by attainment

A quick self-check that keeps ability grouping honest. If any answer is no, use a random draw instead:

1. Is the skill specific? "Weak at maths" is not a grouping criterion; "hasn't secured column subtraction" is.
2. Does it have an expiry date? If you can't say when this grouping dissolves, you're not making a group; you're making a table with a reputation.
3. Would you say the grouping logic out loud? "This group is practising yesterday's skill again" survives being spoken. "This is the low group" does not, and students hear the unspoken version anyway.
4. Is the data fresh? Grouping today's lesson by September's test punishes students for who they were four months ago.

The same test works in reverse when a parent or senior leader asks why your groups are random: "because for this task, mixed groups are the design" is a complete answer, and a visible, live draw means nobody has to take your word for the fairness part.

A practical mixed policy

1. Random by default. Most collaborative tasks don't need engineered groups; they need even sizes, a fair process, and a quick start.
2. Deliberate when targeted. When you group by attainment, tie it to a specific skill, say so plainly, and put an expiry date on it.
3. Never let groups calcify. Whatever the method, the same students in the same groups for a term is the failure mode. Re-deal often.

A grouping tool can make the random part fast, fair, and visibly so, deal the groups live on the projector and the debate about favouritism ends on its own. What no tool can decide is when your class needs design instead of chance. That judgement is the actual teaching.