Why there are three modes, what each one does, and how to choose the right one for your situation.
Clever Groups started with one mode — Preference Groups — and it remains the most widely used. Demand Groups was our most-requested feature: a mode that goes beyond the best overall outcome and guarantees every individual ends up with at least one person they chose. In situations where a good average isn’t enough — camps, overnight trips, transition programs — every individual needs to feel safe and included.
Demand Groups is the direct response: a mode that optimises for each person individually rather than for the overall group average.
Random Groups is for cases where impartiality matters more than preferences — assessment tasks, deliberate mixing, or when no preferences were collected at all.
The three modes cover three genuinely different objectives. Choosing the right one shapes not just the result, but which numbers Clever Groups reports to you afterwards.
Objective: highest overall group satisfaction. The unit of optimisation is the group. Clever Groups builds an arrangement that maximises the weighted average satisfaction score across all groups — earlier preferences count more than later ones.
Preference Groups is willing to accept that a small number of individuals may not get any preference placed, if doing so produces a significantly better result for everyone else. It’s a collective optimisation: the arrangement that does the best overall job of honouring preferences across the whole class or cohort.
The average group satisfaction score is bold — this is the primary metric Preference Groups was optimising for. It reflects the quality of the arrangement as a whole, not a simple count of preferences placed.
Recently we conducted a research project to lift the performance of Clever Groups’ leading preference-based group generation capability. We tested Preference Group Generation’s latest version against more than 600 real group generation exercises from Clever Groups teachers — real class sizes, real preference patterns, real group configurations from years of actual use. Across those exercises — 46,000 student placements in total — the latest version placed around 1,300 more students with at least one person they chose compared to the previous version. That improvement held consistently across every class size tested, from 20-student groups through to 300+ student cohorts.
The same research confirmed Preference Groups’ group satisfaction advantage: it produced average group satisfaction scores around 4–6 percentage points higher than Demand Groups on the same data. Where Demand Groups maximises how many individuals get someone they chose, Preference Groups maximises how well the group as a whole works — the two modes are genuinely different objectives, not one being better than the other.
No student left behind. Demand Groups was built in direct response to the most common request we’ve received: “I need every student to end up with at least one person they know.” Preference Groups finds the best arrangement for the group as a whole — but for camp cabins, overnight trips, and transition programs, “best overall” isn’t the right measure. Demand Groups works to ensure every individual ends up with at least one person they chose, and tells you exactly who couldn’t be placed.
Preference Groups maximises what groups can achieve together — it’s the mode for getting the most out of your class. Demand Groups answers a different question entirely: not “which arrangement produces the highest group satisfaction?” but “does every individual have at least one person they wanted?” It treats at least your first preference as a non-negotiable rather than a wish. Tested across more than 600 real exercises, Demand Groups placed a higher share of students with at least one person they chose than Preference Groups on the same data — that gap is what the mode is built for. The trade-off is a lower group satisfaction score, which is expected: maximising individual placement and maximising what groups achieve together are genuinely different objectives.
Clever Groups considers all the preferences students submitted, but not equally. A student’s first preference carries the most weight, their second preference the second most, and so on — so Clever Groups will always try to place higher preferences first, without doing so at the expense of other students in the team. An arrangement where more students get their top pick is considered better than one where the same number get a lower-ranked preference, even if the raw count of “preferences placed” looks identical.
Students can submit up to N−1 preferences in a team of N students — the more they nominate, the more room Clever Groups has to find a good placement for them.
The stats panel shows you:
Those are the numbers to focus on — clear, actionable, no interpretation needed.
The more preferences each student submits, the more options Clever Groups has when placing them — but only if those preferences point to genuinely different people. A student who nominates only one person leaves nothing to fall back on if that person can’t be placed with them. A student who nominates three different people gives Clever Groups real flexibility.
The caveat: if a student’s preferences are all concentrated on the same highly sought-after people — the few students nominated by half the cohort — those nominations are competing for the same scarce group slots. In a class where ten students all want to be with the same two popular people, most of them won’t get it regardless of how many preferences they submitted. Clever Groups can’t place someone in two groups at once.
The practical advice for teachers: when asking students to submit preferences, encourage them to nominate people they actually want to be with — a friend, someone they know, someone they’d feel comfortable next to on a bus or in a cabin. Demand Groups is about social comfort and belonging, not task output. A mix of nominations that reaches beyond their single best friend gives Clever Groups more room to find a good placement for everyone.
If you’re using Demand Groups and preferences in your class are spread across a reasonably diverse set of people, use the All preferences sub-option — Clever Groups has more to work with. But if nominations are heavily concentrated on the same few popular students, first preference only may produce better results: when second and third preferences mostly point to the same oversubscribed people as first preferences, the extra nominations don’t meaningfully expand the option space, and restricting to first preference gives Clever Groups a cleaner goal — maximise the number of students who get their top pick, full stop.
Two counts appear, and the relevant one is bold depending on which sub-option you ran:
Students who couldn’t be placed (a ✗ glyph) are the ones to focus on first when reviewing manually.
Students are placed into groups by random draw. Preferences are not considered. Any keep-apart pairs you have set are still honoured — those are a hard constraint that applies in all modes.
The same stats appear as for Preference Groups — average group satisfaction score, first preference placed, and any preference placed — computed against the random arrangement. This lets you compare directly: run Preference Groups and note the scores, then run Random Groups on the same team and see what the cohort looks like when placed with no preference logic. The gap is typically substantial — across real exercises, preference-based modes score around three times higher on preference placement than a random draw. That gap is the measurable value of the preference data your students submitted.
| Your situation | Best mode |
|---|---|
| I want the best groups — where students are most likely to engage, collaborate, and produce strong work | Preference Groups |
| I want every student to be with at least one person they chose — individual satisfaction matters more than overall group balance (camp cabins, overnight trips, excursion buddies, transition programs) | Demand Groups |
| I’m using Demand Groups and many students are nominating the same popular few — I want Clever Groups focused on each student’s single top pick rather than chasing fallbacks that point to the same scarce people | Demand Groups — first preference only |
| Rotating groups — I want a different arrangement each session but still preference-aware | Preference Groups (re-run it — the arrangement naturally varies) |
| Assessment tasks requiring impartial, blind grouping | Random Groups |
| No preferences collected — just need groups quickly | Random Groups |
Also see: Understanding your results • Getting started guide • Group sizes, categories & keep-apart rules