Group Exercise: Assessment Centre Practice and Tips

The exercise you can't revise for is the one worth preparing for

Every candidate who walks into a group exercise knows the two pieces of advice: don't dominate, don't go silent. Almost nobody has been told what the space in between actually sounds like. So the room splits into people talking too much, people waiting for a gap that never arrives, and one person who has decided that summarising everyone else's points is a strategy.

Here is the uncomfortable reality that reframes the whole thing: the assessor watching you is not forming an impression. They are holding a sheet with 3 or 4 competencies on it, and under each one, a short list of behavioural indicators. They have 30 minutes to write down enough evidence to defend a score. If you say something excellent but unattributable to an indicator, it does not score. If you say something ordinary that lands squarely on an indicator, it does.

That is why the strongest performers in group exercises are often not the most impressive people in the room. They are the people whose contributions are easiest to write down.

What group exercises measure

The core set is stable across employers: teamwork, verbal communication, listening, influencing without formal authority, and — in many frameworks — leadership potential demonstrated through contribution rather than control. Some employers add analytical reasoning, because the task usually contains real information to handle, and some add time management explicitly.

What assessors record is behaviour, not quality of opinion. The indicators tend to look like this:

• Builds on contributions made by others rather than restating a position
• Invites input from quieter participants
• Manages time or pace on behalf of the group
• Changes position when presented with better evidence
• Summarises accurately without claiming ownership of others' points
• Disagrees without dismissing

Read that list again with the sheet in mind. Every single one of those is a thing you do to other people's contributions. Not one of them is about the strength of your own idea. This is the central misunderstanding: candidates prepare a view, and the exercise is scoring their handling of everyone else's.

There is also a negative column, and dominance sits in it. Talking for 40 percent of a discussion with 5 candidates in it is not scored as leadership. It is scored as a failure to listen, and on most frameworks it costs you 2 competencies at once.

Common formats

Leaderless group discussion. The default. A topic or a decision, 4 to 8 candidates, no assigned roles, typically 5 minutes of reading and 30 minutes of discussion, sometimes with a required output — a ranked list, a single recommendation, or a short summary delivered back to the assessors.

Competitive group task. Everyone gets a shared brief plus a private brief containing information or an objective the others do not have — usually your "own department's" position. Structurally designed to create conflict. The trap is obvious in hindsight and invisible in the room: candidates fight for their private objective and forget that the group is still scored on reaching a decision. Advocating hard for your brief and helping the group land somewhere is the behaviour that scores.

Case-study group task. A substantial information pack — financials, a market summary, 3 or 4 options — analysed collectively under time pressure. More analysis, less social manoeuvring. Common in consulting, banking and public policy.

Timed build or construction task. Physical or simulated, with materials and a spec. Less common now, mostly at in-person centres. Scores planning and role allocation more visibly than the discussion formats do, because the failure modes are physical and immediate.

Assigned-role discussion. Less common, but real: candidates are handed roles (chair, timekeeper, finance lead). Being made chair is not an advantage — chairs are scored on whether everyone got heard, not on whether they ran a tight meeting.

A realistic brief: what you'll actually be handed

Abstract advice is hard to act on, so here is a discussion brief of the type a graduate-scheme candidate genuinely receives. This one is the leaderless format, for a mid-sized UK retail group.

CANDIDATE BRIEF — GROUP DISCUSSION

Time: 5 minutes individual reading. 30 minutes group discussion. 5 minutes group summary to assessors.

Situation: Northgate Retail operates 148 stores. The Board has allocated a one-off budget of £1.2 million to improve staff retention, after turnover among store colleagues reached 38 percent last year against a sector benchmark of 29 percent. Exit interview data suggests the 3 most cited reasons for leaving are unpredictable shift patterns, limited progression, and pay.

Your task: As a group, agree a single recommendation for how the £1.2 million should be spent. You may split the budget across no more than 2 initiatives. You must reach a group recommendation within the time allowed and nominate someone to present it in 5 minutes.

Options under consideration (you are not restricted to these): 1. A 4 percent pay uplift for store colleagues on the lowest grade — consumes the full budget for 1 year. 2. Scheduling software giving colleagues 3 weeks of advance shift visibility — £400,000 build and licence, then £90,000 a year. 3. A structured progression programme with 60 supervisor places a year — £550,000 a year. 4. A retention bonus paid at 12 months of service — £800,000 a year at current headcount. 5. Increased staff discount from 15 percent to 25 percent — £350,000 a year.

Notice what the brief is doing. The budget does not cleanly cover any 2 options. The pay option — the one exit interviews rank third — eats everything. The scheduling option is the only one with a low recurring cost, and it addresses the most cited reason. There is no correct answer, but there is definitely a shape of good reasoning, and the assessors know what it is. There is also a hard time constraint and a required output, which exists so that someone in the group has to manage the clock. That someone can be you, and it is one of the cheapest indicators on the sheet to hit.

A worked excerpt: what a high-scoring contribution sounds like

Below is a short transcript from roughly 14 minutes into a discussion on the brief above, with 5 candidates. Candidate C is the one to watch. Annotations in italics are what an assessor would plausibly be writing.

Candidate A: I still think it has to be pay. Every survey ever done says people leave for money. If we're serious we do the 4 percent and we do it properly.

Candidate B: Agreed. And it's the simplest to implement, we're not building software.

Candidate D: I mean, I don't disagree, but —

Candidate A: It's the only one that's actually material to someone on the lowest grade.

Candidate C: Can I pick up D's point, because I think it got cut off — D, were you going to say something about the recurring cost?

Invites a specific quieter participant back in, by name, referencing their actual point rather than a generic "what do you think?" — meets "invites input from others" and "listens actively."

Candidate D: Yes — the pay uplift is 1 year of budget. There's no money for year 2. So we'd be announcing a rise we can't repeat.

Candidate C: That's the thing that's bothering me too. Three of these five options consume the whole budget annually, so we'd be committing the Board to a permanent cost out of a one-off pot. The scheduling one is the only option where the recurring cost is £90,000 rather than the full amount — and shift unpredictability was the most cited reason in the exit data, not pay. A said pay is what surveys say people leave for, and that's true in general, but this company's own exit data ranks it third. I'd suggest we test the options against the exit data rather than against what we'd each personally want.

Distinguishes one-off from recurring cost — analytical contribution grounded in the brief's actual numbers. Challenges A's position by citing evidence in the brief, without dismissing A. Reframes the group's decision criterion — this is influencing without authority. Meets "changes the basis of discussion using evidence," "disagrees constructively."

Candidate A: Okay, that's fair. I'd still want some pay element in there.

Candidate C: That works — we're allowed 2 initiatives. Scheduling at £400,000 plus the discount increase at £350,000 leaves headroom, or scheduling plus a smaller pay uplift. Can we spend 5 minutes testing those 2 combinations against the 3 exit reasons, and then decide? We're at 14 minutes with 16 minutes left.

Converts disagreement into a concrete next step, offers a compromise that preserves A's position, and names the time explicitly. Meets "manages time on behalf of the group" and "moves the group toward a decision."

Count the contributions. Candidate C spoke 3 times in that excerpt and generated evidence for at least 5 separate indicators. Candidate A spoke 3 times as well, with more conviction and more airtime, and generated evidence for approximately 1 — and one negative note for talking over D.

The mechanics worth copying:

Name people. "Can I pick up D's point" is worth more than the same content said anonymously, because it is evidence of listening rather than a claim about it.

Cite the brief. "Shift unpredictability was the most cited reason in the exit data" is a specific, checkable claim. "I think people care about work-life balance" is an opinion the assessor cannot map to anything.

Say the criterion, not just the conclusion. C's highest-value sentence is "let's test the options against the exit data." That is a contribution about how the group should decide, and it changes the behaviour of 4 other people. Framework contributions score far above position contributions.

Announce the clock once. Not repeatedly — a candidate who reads out the time every 4 minutes is performing timekeeping. Once, at a moment when it changes what the group does next, is enough.

Leave your opponent something. "We're allowed 2 initiatives" lets A move without losing. Assessors score the candidate who makes agreement possible.

How to approach the exercise

In the reading minutes, find the constraint, not the answer. Five minutes is not enough to form a considered view on a case with 5 options. It is plenty to spot that the budget is one-off and 3 of the options are recurring. Candidates who arrive with a constraint are useful in the first 3 minutes. Candidates who arrive with a conclusion are just the first person to state a position.

Contribute inside the first 4 minutes. Not because early speakers score better on merit, but because the longer you wait the higher the bar you set for your own first contribution, and candidates who have been silent for 12 minutes tend to make an oversized entrance to compensate.

Aim for roughly your fair share of airtime, slightly under. In a group of 6 over 30 minutes, that is about 5 minutes of talking. Under-speaking is a real risk and is scored as a gap; over-speaking is scored as a negative. Slightly under fair share, with high-quality interventions, is the optimum.

Handle a dominator by using them, not fighting them. Attacking the loudest person makes the discussion a two-way argument and takes 3 other candidates out of the room, which damages your own listening and inclusion indicators. "That's a strong argument for pay — can we stress-test it against the year 2 cost before we lock it in?" accepts their contribution, redirects the group, and demonstrates influence. It also tends to work.

Protect the output. Many groups produce excellent discussion and no decision. On most sheets, failing to reach a recommendation within the time limit costs every candidate in the room, including the ones who tried to close it — so close it early enough that it survives.

Volunteer to present the summary only if you can do it accurately. It is visible, and it is a genuine chance to demonstrate structured communication. It is also a chance to misrepresent the group's position in front of 5 people who know you did.

How the format shifts by employer type

Graduate schemes run the near-universal leaderless discussion with 5 to 8 candidates, because it scales to a cohort cheaply.

Consulting and finance lean toward case-based tasks with real numbers, and weight analytical contribution more heavily than warmth. The scoring tilts toward whether you handled the data correctly.

Public sector discussions are usually mapped explicitly to a published behaviour framework, and often involve a policy or service-delivery scenario with competing stakeholder interests rather than a commercial one. Use the framework's own vocabulary — evidence that does not map to a named behaviour may not get recorded.

Retail, hospitality and operations employers tend to use practical scenarios — staffing, scheduling, a store-level problem — and weight pragmatism and pace over analytical elegance.

Senior hiring rarely uses group exercises at all, simply because assembling a cohort of directors on one afternoon is impractical. Where they appear, they are usually a facilitated business simulation rather than a discussion.

Common mistakes

Confusing airtime with leadership. Covered above, and still the single most common failure. Dominance is scored negatively on most modern frameworks.

Restating instead of building. "I agree with what Sarah said, I also think we should consider cost" is scored as a contribution of roughly zero. "Building on Sarah's point about cost — the recurring versus one-off distinction means option 2 is the only sustainable one" is scored as 2 indicators.

Treating other candidates as rivals. Groups that visibly compete produce worse discussions, and everyone's score drops together. All the evidence suggests the room is scored against a standard, not ranked against itself.

Chairing without being asked. Taking charge of process in the first 60 seconds of a leaderless discussion reads as a prepared gambit, and often triggers a second candidate to contest it. Process contributions land much better at minute 12 than at minute 1.

Going silent after being contradicted. Being argued down is not a scoring event. Withdrawing from the discussion afterwards is.

Ignoring the clock entirely. In a 30 minute discussion with a required output, the group that reaches minute 25 with no recommendation has failed a shared objective.

How to prepare

1. Practise with 3 or 4 other people, not alone. Any group. Give yourselves a brief like the one above, set a 30 minute timer, and have someone observe silently and count how many times each person built on somebody else's point. The number is usually lower than participants think.
2. Write 5 reusable sentence forms and rehearse them out loud. The invite ("D, you were about to make a point about cost — can we come back to it?"), the build, the evidence-based challenge, the criterion proposal, the time-check. These are not scripts for content; they are handles for behaviour under pressure.
3. Read the employer's competency framework and translate it. For each indicator, write the sentence that would demonstrate it. That translation step is what most preparation skips.
4. Get comfortable being wrong out loud. "That's a better argument than mine, I'd change my position" is an explicit indicator on many frameworks and almost nobody says it.
5. Practise the 5 minute summary. Take any 20 minutes of discussion you have been part of and summarise it in 90 seconds, accurately, attributing points to the people who made them.
6. For online centres, rehearse the interruption problem. Video call latency of even 200 milliseconds destroys natural turn-taking. Practise the raised hand, the short verbal signal, and the deliberate half-second pause before speaking.

Frequently asked questions

TestSolve is independent and not affiliated with any test provider or employer named on this page. All product names and trademarks belong to their respective owners.