SBI PO Group Exercise 2026: Format, 20 Marks, a Full Worked GE and 30 Topics

What the Group Exercise is, per SBI

The SBI PO Group Exercise is the first spoken part of Phase III and is worth 20 marks, against 30 for the personal interview, per the SBI PO 2026-27 detailed advertisement (CRPD/PO/2026-27/09). The advertisement says Phase III "comprise of Psychometric Test, Group Exercise & Personal Interview", gives the two maximum marks in a table, and adds that "the minimum qualifying marks for Group exercise & Interviews will be as decided by the Bank". The 50 Phase III marks are normalised to 25 of the final 100, with the main examination's 230 normalised to 75. Per the SBI Phase III notice, the Group Exercise and Interviews run from 18 November 2026, after a psychometric test on 31 October 2026. That is everything SBI publishes about the GE. Everything else on this page, the group size, the timing, the task types, comes from what candidates report, and we say so where it does.

Two consequences follow from the official text. First, because Phase III is normalised from 50 to 25, one Group Exercise mark is worth half a final mark, while one main examination mark is worth 75 divided by 230, about 0.33. A GE mark is worth about one and a half mains marks. Second, because the qualifying marks are "as decided by the Bank" and a candidate must qualify in Phase III separately, a weak GE is not simply a few marks lost; it can be the reason a strong mains score never reaches the final list. The arithmetic, with worked examples, is on our marks and weightage page.

What candidates report about the format

SBI's advertisement does not describe how the Group Exercise is run. The pattern below is assembled from candidate accounts of recent Phase III rounds. It is consistent across accounts, but it is not a rule, and the officers on the day can change any part of it.

Three task types recur in these accounts. The ranking task: "Here are eight measures to improve financial inclusion; as a group, rank the top three." The case task: "A branch has a queue problem and a budget; agree what to do." The debate-style topic: "Should public sector banks be privatised?" where the group is still asked to reach a position. Prepare for all three, because the task sheet decides which roles are available to you.

How a GE differs from a GD

Most candidates prepare for the SBI Group Exercise with group discussion material from MBA admissions, and most of that material rewards the wrong behaviour. A GD is scored by how well you argue. A Group Exercise, as SBI's name and candidates' accounts both suggest, is scored by how well the group gets something done and what you contributed to it. The table sets out the difference.

The practical shift: in a GD you prepare arguments; in a GE you prepare moves. The moves are the same for every topic: propose a structure, add a point with a reason, build on someone, resolve a clash, bring in a quiet member, summarise. The group discussion phrases post lists the English for each move; the section below adapts them to a banking GE.

The roles that score, and the ones that do not

Observers cannot score twelve people on twelve dimensions in twenty minutes. Candidates who have sat the GE and officers who have run assessment exercises describe the same shortcut: each observer notes, for each candidate, two or three moments that fit a recognisable role. Make sure you give them good moments to note.

Roles that score

• The structurer. Opens not with an opinion but with a plan: "We have about fifteen minutes and three options to rank. Shall we take two minutes to hear one point each, then rank?" Almost nobody does this; the one who does is remembered as the leader without having claimed it.
• The builder. Adds to someone else's point by name or number: "Adding to what number six said about UPI, the same logic applies to..." Observers hear the name and tick the box for listening.
• The evidence-giver. Brings one concrete example per turn: a scheme, a branch situation, something from their own district. Not statistics they cannot source; a panel officer may know the real figure.
• The resolver. When two members lock horns: "I think you are both right about different things. Number three is talking about cost and number nine about reach. Can we agree reach matters more for this task and move on?"
• The includer. Brings in a member who has not spoken: "Number eleven, you mentioned rural branches earlier, what is your view on option two?" One such move is worth more than three extra points of your own.
• The summariser. With two minutes left: "Let me try to sum up where we are. We agree on A and B; we are split on C; shall we go with B first?" The person who closes the group is the person the observers remember last.

Roles that do not score, or score negatively

• The dominator. Speaks most, interrupts, repeats. Fluent and scored down. Observers are watching the group's faces as much as the speaker.
• The statistician. Quotes numbers nobody can check. One wrong number noticed by an officer costs more than ten right ones gain.
• The silent member. Speaks once or not at all. Not necessarily a zero, but the observer has nothing to write.
• The echo. Repeats what was just said in different words to be seen to speak. Observers notice.
• The debater. Treats the GE as a GD and tries to win. Wins the argument; loses the role.
• The off-topic expert. Pulls the task towards the one thing they know. Scores for content, loses for the group outcome.

If English is the worry: the structurer, the includer and the summariser roles need the fewest words. Each is one or two sentences that can be prepared in advance. A Hindi-medium candidate who opens with the structuring line and closes with the summary line, and speaks twice in between, usually outscores a fluent dominator. The phrases are in the section below and the L1-specific fixes are on our Hindi speakers page.

A full worked GE, written by TalkDrill

The transcript below is written by TalkDrill to show the moves in sequence. The eight participants are invented and numbered as they would be in the room. The task is a ranking task of the kind candidates report. It is shortened: a real GE of fifteen to twenty minutes would have about twice as many turns. Read it once for the content, then a second time noting only who makes which move.

Participant 2 (structurer): "Good morning, everyone. We have fifteen minutes and six options. May I suggest we spend the first four or five minutes hearing one point from each person, then spend the rest narrowing to three? And perhaps we keep the last two minutes for the summary." (Several nod. Nobody objects. Participant 2 has led without claiming to.)

Participant 5: "Agreed. I will start. I think (d), the vernacular app with voice, has to be first. In my district most people over forty will not touch an English app, and voice guidance removes the reading barrier entirely. It is a one-time cost that then serves every branch."

Participant 1: "I would put (f), training at SHG meetings, very high. SHG members already trust the bank, they meet every week anyway, and in my experience when the women in a group adopt something the whole village follows. It costs almost nothing."

Participant 7 (builder): "Building on what number one said, (f) and (c) work together. The business correspondent is often the person who attends the SHG meeting. If we incentivise the BC and give them the SHG platform, we get the training and the onboarding in one move. So I would rank (c) and (f) as a pair."

Participant 3: "I disagree with putting the app first. An app nobody has been shown is just a download. I would fund (a), the help desk, first, because it is where the older customer actually goes. They walk in, someone sits with them for ten minutes, and they leave with UPI working."

Participant 8: "The cashback, (e), is what the private players did and it worked. People try things for money. I would keep it in the top three."

Participant 4 (evidence-giver): "On (e), one caution. Cashback brings transactions while it runs and they fall when it stops; I have seen that with shopkeepers in my town who switched apps every month for the offer. The task says raise transactions without losing older customers. I would rank (e) low, because it does not address the older customer at all. On (b), doorstep banking, I think it is valuable but expensive per customer; one staff member once a week reaches perhaps a dozen homes."

Participant 6: "I agree with number four on (e). On (b), it may matter more for pensioners and the disabled than the numbers suggest, but I accept it is not a top-three measure for digital adoption."

Participant 2 (structurer, keeping time): "That is everyone once. We seem to have agreement that (e) is low and (b) is low. The live options are (a), (c), (d) and (f), and number seven has paired (c) with (f). Can we take two minutes on (a) versus (d), which is the real disagreement?"

Participant 5: "I will defend (d). The help desk serves one branch; the app serves all of them. And a voice app is itself a help desk in the customer's pocket."

Participant 3: "And I will defend (a). The task is about rural and semi-urban branches, not the whole bank. Adoption happens face to face there. The app is necessary but not sufficient."

Participant 7 (resolver): "I think you are both right about different things. Number five is right that (d) is the biggest single lever; number three is right that nothing gets adopted without a human first step. What if we rank (d) first because it is the enabler, and (a) second because it is the delivery, and we note in the summary that they are funded together?"

Participant 5: "I can live with that."

Participant 3: "So can I, if the summary says they go together."

Participant 1 (includer): "Number eight, you argued for cashback earlier. Given where we are, would you accept (c) plus (f) as the third measure, since the BC incentive is also a kind of incentive, just paid to the onboarder rather than the customer?"

Participant 8: "Yes, that is fair. Paying the BC per successful onboarding is cheaper than paying every customer, and it lasts."

Participant 6: "One addition for the summary: we should say we would review after six months and consider (b) for pensioners if the numbers show older customers dropping off. That answers the second half of the task."

Participant 4: "Agreed. That protects the older customer, which is the constraint."

Participant 2 (summariser): "We have about two minutes. Let me try to summarise and please correct me. First, (d), the vernacular voice app, as the enabler. Second, (a), a digital help desk in each branch, funded together with (d) as the human first step. Third, (c) and (f) together: incentivise business correspondents and deliver the training through SHG meetings. We rank (e) cashback low because it does not serve the older customer and its effect fades. We rank (b) doorstep banking low on cost per customer but recommend a six-month review for pensioners. Is that the group's position?" (Nods around the circle.) "Then shall number six present, since the review point was theirs?"

Participant 6: "Happy to." (Presents the ranking in about forty seconds, in the order agreed.)

What the observers would have written

Notice what nobody did: nobody quoted a statistic, nobody interrupted, nobody spoke for more than about thirty seconds at a stretch, and the group reached a ranked answer with a reason for each rank. That is a scoring GE. Run the same task on our group discussion practice with AI participants and try to take participant 2's role.

Thirty topics to practise

These thirty cover the three task types candidates report. Practise each as a GE, not a GD: set a task (rank, decide, recommend), set fifteen minutes, and end with a summary. For every topic, know two concrete examples and no unsourced figures. The banking awareness page gives the concepts behind the banking topics in plain English.

Banking and finance (ranking or recommendation tasks)

1. Rank five measures to reduce NPAs in a mid-sized public sector bank.
2. Should public sector banks be privatised? Reach a group position with two conditions.
3. Rank six ways to deepen financial inclusion in a hill district.
4. Digital-only banks versus branch banking: recommend a model for semi-urban India.
5. How should a bank respond to UPI fraud complaints rising at its branches? Agree three steps.
6. Should the bank lend more to agriculture or to small manufacturing? Decide for one district.
7. Rank five uses of artificial intelligence in a bank by benefit to the customer.
8. Should bank branches be open on Sundays? Reach a recommendation.
9. A branch has a queue problem and a small budget. Agree a three-point plan.
10. Rank four groups for a financial-literacy drive: students, SHG women, pensioners, shopkeepers.
11. Cash versus digital: should the bank discourage cash at the counter? Decide with safeguards.
12. Green lending: rank five sectors a bank should prioritise for climate-linked loans.
13. Should banks sell insurance and mutual funds at the counter? Agree limits.
14. Cryptocurrency and banks: recommend how a bank should treat customers who trade in it.
15. Central bank digital currency: rank three use cases for a rural branch.

Economy and society (debate topics that still need a conclusion)

16. Is a cashless economy desirable for India? Agree a position and one risk.
17. Gig work: should platform workers get social security through their bank accounts? Recommend a mechanism.
18. Work from home for bank staff: which roles, if any? Decide.
19. Should higher education be financed mainly by loans? Reach a position.
20. Women's participation in the workforce: rank four measures a bank could support.
21. Free public transport in cities: should banks finance it? Agree.
22. Rural to urban migration: is it a problem to solve or a trend to serve? Decide for banks.
23. Social media and bank reputation: agree three rules for staff.
24. Electric vehicles: should banks offer cheaper loans for them? Recommend.
25. Data privacy: how much customer data should a bank use to sell products? Agree a line.

Situations and judgement (case tasks)

26. A large customer asks a new PO to bend a know-your-customer rule. Agree what the PO should do, step by step.
27. A branch must cut costs by a fixed share. Rank five options and name what you would not cut.
28. A village wants an ATM; the bank's data says it is not viable. Agree a response.
29. Ten people in a lifeboat for six: as a committee, agree criteria without naming people.
30. Your group has a fixed sum to spend on one of three community projects: a school computer lab, a water tank, or a market shed. Decide and justify.

Several topics have ready-made practice rooms: topic 7 at AI and jobs, topic 17 at the gig economy, topic 21 at free public transport, topic 24 at electric vehicles and topic 25 at data privacy. The closest banking case room is a bank app's falling ratings, which runs like topic 9. Our placement-season list of group discussion topics has forty more if you run out.

Phrases that carry a turn

Each move has a sentence that does most of the work. Learn one or two per move until you can say them without thinking, so your attention goes to the content. These are adapted for a banking GE from our fuller list of group discussion phrases.

Two phrases to drop: "I would like to add my valuable point" (nobody's point is valuable until it is heard) and "As we all know" (if everyone knows it, do not say it). And one habit to watch for Hindi, Bengali, Telugu, Tamil, Marathi and Odia speakers alike: the tag "no?" at the end of a sentence ("UPI is better, no?") reads as unsure in a GE. End the sentence instead.

How to practise it before 18 November

The Group Exercise cannot be prepared by reading. It needs other voices, a timer and feedback. Here is the route we recommend between now and your call-letter date; the dated version sits inside our three-week plan.

1. Week of 20 October: learn the moves. Read the worked GE above twice. Write your own version of each phrase in the table in words you would actually say. Read the mock group discussion with AI guide to see how a practice round is scored.
2. Week of 27 October: one GE a day, fifteen minutes. Pick a topic from the thirty. Run it on group discussion practice, where AI participants take the other seats and the transcript shows who spoke when. Your target for the week: open with the structuring line in at least three rounds.
3. Week of 3 November: roles on rotation. Each day, aim for one role you did not take yesterday: builder, resolver, includer, summariser. Read the transcript afterwards and count your turns; three to five good ones in fifteen minutes is the aim, not ten.
4. Week of 10 November: pair with people. If you can find four or more candidates, run live GEs with one person observing and using the table above as a scoresheet. Rotate the observer. The AI rounds have taught you the moves; live rounds teach you to make them while someone is talking over you.
5. Interview week: taper. One GE every second day, no new topics, and read the SBI PO interview page for what happens on the day. The GE and the interview are on the same day by every account; a calm GE puts you in the interview room already warmed up.

For the personal interview that follows the GE, start with your introduction, since panels at SBI often open with "we saw you in the group exercise; tell us about yourself". The Bank PO interview hub carries the live dates block, which is updated from the SBI notice and the IBPS calendar each week.