Group discussion for analytics roles: a sample round, scored

A simple rubric to score yourself

Use a 1–3 scale on each line. Evaluators' actual forms vary, but most cover these ideas.

A total of 11 or more across five criteria is a strong performance on this scale.

The sample round

Topic: "Should colleges use data to predict which students will fail and intervene early?" Four fictional participants — Arjun, Megha, Farhan and Neha — have ten minutes.

Arjun: "I'll start. Data is the future and every college should use it. Companies use data for everything, so colleges should too."

Score: content 1, structure 1, evidence 1. It is a slogan, not a reason. "Companies do it" does not show that it helps students.

Megha: "I think there are two questions here: does prediction actually help students, and is it fair to them? On the first, attendance and internal marks are available early in the semester, so a college could flag students a few weeks before the final exam, when there is still time for extra classes."

Score: content 3, structure 3. She frames the discussion into two parts and gives a concrete mechanism.

Farhan: "But what if the model is wrong? A student labelled 'likely to fail' may be treated differently by teachers."

Score: content 2, listening 2. A real risk, raised clearly. It would score higher by linking to Megha's second question explicitly.

Arjun: "Yes, but data is always better than gut feeling."

Score: listening 1. He does not answer Farhan's point.

Neha: "Building on Farhan's point: the fairness risk depends on what the college does with the flag. If it only offers optional extra help, a wrong flag costs a student an invitation they can ignore. If it changes grades or seat allocation, a wrong flag costs much more. So the design of the intervention matters more than the model."

Score: content 3, listening 3, structure 3. She connects two earlier points and moves the group forward.

Megha: "Farhan, you raised it — would you accept the system if the only action is an optional tutoring invitation?"

Score: group contribution 3. She invites a response instead of taking the floor again.

Farhan: "Yes, if students can see why they were flagged and the college checks how often the flags are wrong."

Neha (summarising near the end): "So we agree prediction can help if it catches students early, the action is supportive rather than punitive, students can see the reasons, and the college measures how often it is wrong. Where we differ is on how much data is enough to act on."

Score: group contribution 3. A fair summary that includes the disagreement, not only the agreement.

What the round teaches

• Arjun spoke first and often but scored lowest. Volume without reasons does not help.
• Megha used a framework in her first turn and later brought someone in.
• Neha never opened, but she connected points and summarised, which evaluators remember.
• Nobody invented a statistic. That was correct: none of them had a source, and a made-up "70% of students" figure is easy to challenge and damages credibility.

Bringing numbers in without bluffing

Analytics evaluators notice how you use numbers. Three honest ways:

1. 1Cite what you actually know, with the source: "NPCI publishes monthly UPI transaction statistics, and last month's figure was…" — only if you have actually read that month's figure.
2. 2Label an estimate: "Suppose a college has 2,000 students and flags 10%. That is 200 students; if half of those flags are wrong, 100 students get an unnecessary invitation."
3. 3Ask what data would settle it: "To decide this, I would want to know how often last year's flags were correct."

Option 2 is often the strongest in an analytics GD because it shows you reason with numbers without claiming facts you do not have.

Topic bank for practice

• Should UPI transactions carry a small fee to fund the system?
• Will generative AI reduce entry-level analyst jobs, or change them?
• Should social media platforms be required to verify users' ages?
• Is work from home better for freshers or experienced employees?
• Should cities price parking by demand?
• Do exam rankings help or harm learning?

Practise with three friends, one acting as evaluator with the rubric above. Rotate roles so everyone scores and is scored. Record the round on a phone if everyone agrees, and watch where you interrupted or repeated.

If the group is chaotic

When two people talk over everyone, wait for a pause and enter with a summary: "Let me pull together what we have so far — two points on cost and one on fairness." Summarising earns you the floor because it is useful to everyone. Never raise your voice to compete.

GD often sits in the same selection day as a technical interview; the placement plan for non-CS branches shows how to prepare both. Reasoning out loud with labelled estimates, as Neha and Megha did, is also how analysts present findings to managers — a skill the Data Analyst programme practises in every project review.