Accenture Interview Questions and Answers

Answer

As of 2026 the process candidates typically report has three phases. Phase one is registration and eligibility screening, usually through the Accenture careers portal or a campus drive, where your degree, branch, graduation year, aggregate percentage and backlog status are checked before you get a test slot. Phase two is the online assessment, which is not one paper but a sequence: a cognitive assessment covering verbal ability, critical reasoning and abstract reasoning, a technical assessment covering common applications and MS Office concepts, pseudocode, networking, security and cloud fundamentals, a coding section, and a communication assessment that records and scores your spoken English.

Phase three is the interview stage, normally a technical interview, then a managerial or client facing round, then HR. Some drives merge the technical and managerial rounds into a single longer conversation, and role specific drives sometimes add a domain round. Exact section counts, timers and cutoffs change between drives, so treat any fixed number you read on a forum as indicative rather than official.

What the interviewer is probing when they ask you this at the start of a round is whether you researched the company at all. A weak answer sounds like 'first there is an aptitude test and then interviews', which tells them you registered without reading anything. A strong answer names the communication assessment specifically, because that is the stage most candidates do not know about.

Answer

A normal campus aptitude paper is mostly quantitative: percentages, time and work, profit and loss, permutations. The Accenture cognitive assessment leans harder on language and pattern recognition. Verbal ability covers reading comprehension, sentence correction, para jumbles, error spotting and vocabulary in context.

Critical reasoning covers statement and conclusion, assumptions, strengthen or weaken an argument, and syllogisms, which are reasoning questions rather than arithmetic. Abstract reasoning covers non verbal series, odd one out, figure matrices and analogies, where you infer the transformation rule from shapes with no words at all. Quantitative content still appears in the wider assessment, but candidates consistently report that the surprise is how much of the score sits in verbal and non verbal reasoning rather than in maths.

What this means practically is that a candidate who only practised R S Aggarwal quantitative chapters is preparing for a different exam. The interviewer or the recruiter is not asking you this in a room, but the framing matters because it determines where your revision hours go. The weak preparation pattern is spending three weeks on quant shortcuts and thirty minutes on abstract reasoning the night before, then losing the section that has the least forgiving time per question. Give abstract reasoning and critical reasoning at least equal weight, and treat verbal ability as a scoring section rather than a filler one.

Answer

Proctored assessment platforms commonly lock each section to its own timer, and many do not allow you to return to a completed section, so treat each block as a standalone paper with its own budget. The tactic that works is a two pass sweep inside every section. Pass one, move fast and answer only what you recognise within a few seconds, marking anything that needs working out.

Pass two, return to the marked ones with whatever time remains. This protects you from the classic failure where a candidate spends four minutes on question three, panics at question twenty, and leaves eight easy questions unattempted. Decide your abandon rule before the test starts, for example, any single question that crosses ninety seconds gets a best guess and a mark.

Also decide your guessing policy in advance based on whether the drive announces negative marking. If negative marking is not announced, leaving a blank is strictly worse than a considered guess. What an interviewer probes if this comes up in conversation is whether you are systematic under pressure or whether you improvise.

A weak answer is 'I just try to be fast'. A strong answer names a specific abandon threshold and a specific two pass method, because that is a repeatable process rather than a hope.

Answer

Ten days is enough if you allocate by gate rather than by comfort. Days one to three, cover the cognitive side: one hour verbal ability, one hour critical reasoning, one hour abstract reasoning daily, all timed. Days four to six, cover the technical assessment surface: pseudocode tracing, basic networking, security fundamentals, cloud fundamentals, and common applications and MS Office concepts.

Days seven and eight, coding practice in one language, focusing on arrays, strings, loops, and simple search or sort logic with clean input handling, not on advanced algorithms. Day nine, the communication assessment, which most candidates skip entirely: record yourself reading passages aloud, repeating sentences and retelling short stories, then listen back. Day ten, one full length timed mock and rest.

The habit that fails is spending eight of the ten days on the thing you already enjoy, usually coding, and arriving at the abstract reasoning block cold. A second failure is preparing only content and never rehearsing the mechanics: the platform, the timer, the microphone, the network. If you are asked in an HR round how you prepared, this answer doubles as evidence of planning, and a weak version sounds like 'I revised whatever I could find', which signals no prioritisation.

Answer

The recurring shapes are reading comprehension with inference questions, sentence correction, para jumbles, error spotting, fill in the blanks with contextual vocabulary, and synonym or antonym in context. Speed comes from method rather than from reading faster. For reading comprehension, read the questions first, then scan the passage for the relevant paragraph, because most questions are locatable rather than holistic.

For para jumbles, find the opening sentence by elimination, since an opener rarely starts with a pronoun, a connector like 'however' or 'therefore', or a definite reference to something not yet introduced, then chain the pairs. For error spotting, run a fixed checklist in order: subject verb agreement, tense consistency, preposition usage, article usage, pronoun reference, parallel structure. A checklist beats intuition because intuition trained on spoken Indian English will accept several constructions that the test marks wrong, for example 'discuss about', 'return back', or 'one of the student'.

For vocabulary, learn words in collocation rather than as isolated meanings, because the questions test usage in context. The weak preparation pattern here is memorising word lists, which produces recognition without usage, and then failing exactly the fill in the blanks questions the list was supposed to help with.

Answer

Break the argument into three parts before you look at the options: the conclusion, the evidence offered for it, and the unstated assumption connecting them. Almost every strengthen or weaken answer operates on the assumption, not on the evidence. If the argument is 'the new onboarding flow raised signups by 20 percent, so the redesign works', the assumption is that nothing else changed in that period.

An option saying a large advertising campaign ran the same week weakens it, because it attacks the assumption. An option saying the redesign also increased time on page is irrelevant, because it does not touch the causal link. Two habits protect your score.

First, ignore your own real world knowledge; the test scores what follows from the passage, not what is true in the world, and candidates lose marks arguing from outside information. Second, watch quantifier traps: 'all', 'most', 'some' and 'only' change what a conclusion can legitimately claim, and a valid sounding option that overreaches from 'some' to 'all' is wrong. The weak answer pattern is picking the option that simply agrees with the conclusion in different words, which restates rather than strengthens. Name the assumption out loud in practice; it converts a guessing exercise into a mechanical one.

Answer

Abstract reasoning tests whether you can infer a rule from visual information with no language attached, which is a proxy for learning a new system quickly. The question forms are figure series, figure matrices, odd one out, and analogies. Improvement comes from having a fixed inspection order rather than staring at the whole image.

Check rotation first, then reflection, then movement of an element by a fixed number of positions, then addition or removal of elements, then a count that increases or decreases, then shading or fill changes, then size changes, then a swap of position between two elements. Most items combine exactly two of these rules, so once you identify one transformation, immediately hunt for the second rather than committing to an answer. The other reliable technique is elimination on a single feature: if the rule says the count of dots increases by one each frame, discard every option with the wrong dot count before you consider anything else, which usually leaves two candidates.

What fails is treating each item as a fresh puzzle with no method, which is why untrained candidates find this the most time expensive section despite it having no content to learn. Twenty practice items a day for a week produces a visible jump, because you are building a checklist, not knowledge.

Answer

This block surprises engineering candidates because it is not programming. It tests whether you can operate the tools a delivery team uses daily. Expect questions on spreadsheet behaviour, for example what an absolute reference like $A$1 does when you drag a formula, what VLOOKUP or XLOOKUP returns when there is no match, what a pivot table is used for, and the difference between sorting a range and sorting a table.

Expect word processing questions on styles, headers and footers, track changes and mail merge, and presentation questions on slide masters and animations. Expect general computing questions on file formats, keyboard shortcuts, printing and page setup, email etiquette fields such as CC versus BCC, and basic operating system concepts like file paths and extensions. None of this is hard, which is exactly the trap: candidates skip it entirely because it feels beneath them and then drop easy marks.

Half a day of revision covers the whole surface. If a version of this comes up in an interview, the interviewer is probing whether you will need hand holding on day one at a client site where the deliverable is a spreadsheet or a status deck, not a repository. A weak answer is 'I know Excel', with no ability to name a lookup function or explain a pivot.

Answer

The depth is conceptual, not certification level. On networking, know the OSI and TCP/IP layers and what lives at each, TCP versus UDP and when you would choose each, what an IP address, subnet mask and gateway do, DNS resolution, DHCP, HTTP versus HTTPS, common ports, and the difference between a router, a switch and a hub. On security, know the CIA triad, authentication versus authorisation, symmetric versus asymmetric encryption, hashing versus encryption and why passwords are hashed and salted, what a firewall does, what phishing and social engineering look like, and named application risks such as SQL injection and cross site scripting.

On cloud, know IaaS, PaaS and SaaS with an example of each, public versus private versus hybrid, what virtualisation and a hypervisor are, what elasticity and scalability mean, and roughly what compute, storage and identity services do. Provider specific naming for AWS, Azure and Google Cloud helps but nobody expects you to recite service catalogues. The interviewer is probing whether you can hold a sensible conversation with a client architect, not whether you can configure a VPC. A weak answer is defining hashing as 'a type of encryption', which is the single most common fundamentals error and immediately signals memorised notes.

Answer

Pseudocode questions are deliberately written in a language agnostic style with keywords like BEGIN, SET, FOR, IF, ENDIF and PRINT, because the test wants to know whether you can read logic without an IDE, a debugger or syntax colouring. The reliable method is a written trace table. Draw one column per variable plus one for the loop counter, then write a new row for every iteration, updating only the variables that change.

Never hold two variables in your head. The two traps that appear again and again are order of assignment, where a variable is used on the right hand side after it was already updated earlier in the same iteration, and a condition that compares against a variable which is itself being mutated inside the loop, so the comparison target moves. Both are invisible if you trace mentally and obvious if you write rows.

The worked example below shows the table for a loop where both the accumulator and the comparison target change. What the interviewer probes if they hand you pseudocode live is your process, not just the number, so narrate the trace out loud row by row. A weak answer is silence followed by a guessed number, because even when the number is right the panel learns nothing about how you got it, and they will assume you pattern matched.

BEGIN
  SET a = 4
  SET b = 7
  SET c = 0
  FOR i = 1 TO 3 DO
    c = c + (a * i)
    IF c > b THEN
      b = b + c
      c = c - 2
    ENDIF
  ENDFOR
  PRINT b, c
END

TRACE TABLE
i    c after line 1      c > b ?      b after      c after
1    0 + 4 = 4           4 > 7  no    7            4
2    4 + 8 = 12          12 > 7 yes   7 + 12 = 19  12 - 2 = 10
3    10 + 12 = 22        22 > 19 yes  19 + 22 = 41 22 - 2 = 20

OUTPUT: 41 20

WHY PEOPLE GET THIS WRONG
1. They compare c against the original b = 7 in every iteration,
   forgetting b is mutated inside the loop.
2. They apply c = c - 2 even when the IF did not execute.

Answer

Candidates consistently report that the coding section sits at an easy to moderate level rather than a competitive programming level. The recurring shapes are string manipulation such as reversal, palindrome checks, vowel or character counts and anagram checks, array work such as finding maximum, minimum or second largest, frequency counting, removing duplicates and simple two pointer problems, number logic such as prime checks, factorial, Fibonacci, Armstrong numbers and digit sums, basic pattern printing, and occasionally a simple sorting or searching implementation. Time and space complexity is usually not the deciding factor, correctness on hidden test cases is.

Choose the language you can write correctly without documentation, which for most Indian candidates is C, C++, Java or Python. Java and Python are the safest picks because input parsing is less error prone and standard library support for strings and collections is stronger. Do not pick a language because it looks impressive; a candidate writing shaky C++ pointer code will lose to a candidate writing plain Java.

The failure mode that costs the most marks is not algorithmic, it is input and output handling: reading the wrong number of values, printing extra text around the answer, or a trailing newline mismatch. Practise reading input from standard input and printing exactly what the problem asks, nothing more.

Answer

This is a representative coding round problem: trivial to state, easy to get wrong on hidden tests. The naive approach is to sort and read the second element from the end, which is O(n log n) and, more importantly, still wrong on duplicates because the second element from the end of [9, 9, 4] is 9, not 4. The correct approach is a single pass with two trackers, first and second, updating first when a larger value arrives and pushing the old first down, and updating second only when the value is strictly less than first and strictly greater than second.

The edge cases that hidden tests reliably include are N equal to one, all elements identical, arrays containing negative numbers only, and values that overflow a 32 bit integer. The negative case is where most submissions die, because candidates initialise first and second to zero instead of to the minimum possible value, so an array of [minus 5, minus 2, minus 9] returns zero. What the interviewer probes here is whether you enumerate edge cases before writing code or after failing tests.

A weak answer starts typing immediately and then patches. A strong answer says the four edge cases out loud first, then writes code that already handles them, then dry runs it on the smallest failing input.

import java.util.Scanner;

public class SecondLargest {
  public static void main(String[] args) {
    Scanner sc = new Scanner(System.in);
    int n = sc.nextInt();
    long first = Long.MIN_VALUE;
    long second = Long.MIN_VALUE;

    for (int i = 0; i < n; i++) {
      long x = sc.nextLong();
      if (x > first) {
        second = first;
        first = x;
      } else if (x < first && x > second) {
        second = x;
      }
    }

    if (second == Long.MIN_VALUE) {
      System.out.println("NA");
    } else {
      System.out.println(second);
    }
  }
}

EDGE CASES THE HIDDEN TESTS USE
n = 1                    => print NA
all elements equal       => print NA, distinct is the key word
all values negative      => must not seed trackers with 0
values beyond int range  => read into long
Extra text in output     => print the number only, no labels

Answer

Sample cases are the happy path; the score comes from hidden cases that target boundaries. The usual causes, ranked by how often they occur. One, unhandled edge inputs: empty input, a single element, all elements identical, all negatives, or the maximum stated N.

Two, integer overflow, where a sum or product exceeds a 32 bit type and silently wraps rather than erroring. Three, output formatting drift: printing 'The answer is 41' when the grader expects '41', printing on multiple lines when one is expected, or trailing whitespace differences. Four, input parsing that assumes a format the problem never promised, for example assuming values are space separated on one line when they arrive one per line, or reading N values when the actual count is on the previous line.

Five, time limit exceeded from a nested loop over a large N when the constraints implied a linear solution. Six, off by one errors on loop bounds, which produce right looking answers on small samples and wrong answers on the largest case. The disciplined fix is a personal checklist you run before submitting: smallest input, largest input, all identical, all negative, exact output format. What separates a strong candidate is treating a low score as a diagnosable defect with a known list of causes, rather than as bad luck.

Answer

The communication assessment is a separate stage where your spoken English is recorded and scored by an automated system rather than by a person. Candidates report task types including sentence repetition, where you hear a sentence and repeat it exactly, reading passages aloud, story retelling, where you listen to a short passage and then narrate it back in your own words, and sometimes short open ended speaking prompts. The system evaluates measurable properties of speech: pronunciation clarity, fluency and pace, sentence stress and rhythm, vocabulary and grammar in the free speech tasks, and how completely you reproduce content in the repetition and retelling tasks.

Technically strong candidates fail it for three reasons. They did not know it existed, so they never rehearsed speaking aloud. They speak too fast because they are nervous, which destroys the clarity score even when the English is grammatically perfect.

Or they treat a low grade microphone and a noisy room as irrelevant, when the system is scoring the audio it receives, not the sentence you intended. Treat this as a performance stage, not a knowledge stage. A weak preparation approach is reading grammar rules. The right preparation is recording yourself, listening back, and fixing pace and articulation.

Answer

These two tasks reward different things, so practise them separately. Sentence repetition tests working memory for language plus articulation. The technique is to listen to the whole sentence for meaning rather than trying to capture individual words, because a sentence you understood is far easier to reproduce than a string you memorised.

Do not begin speaking before the audio ends, and do not pause mid sentence to search for a word; a fluent approximation of the full sentence usually scores better than a stalled exact one. Practise with progressively longer sentences, from eight words up to about twenty, and repeat immediately without replaying. Reading aloud tests pronunciation, pace, phrasing and stress.

Practise on unfamiliar text, since reading something you have memorised hides exactly the problems the test finds. Mark natural pause points at commas and clause boundaries, keep pace steady, and finish every word, because dropped final consonants are the most common clarity issue for Indian candidates, for example 'projec' for 'project' or 'developmen' for 'development'. Record, listen, and note the three sounds you personally distort, then drill only those.

What fails is silent preparation. You cannot fix a speaking score by reading; the feedback loop only exists when you hear your own recording.

Answer

Story retelling scores how much of the content you reproduce and how coherently you structure it, so go in with a fixed skeleton rather than trying to remember sentences. While listening, hold five slots in your head: who, where, what happened first, what changed, and how it ended. Do not attempt to memorise wording; the task explicitly asks you to retell in your own words, and paraphrasing scores fine as long as the facts and sequence survive.

When you speak, open with a one line frame, then narrate in chronological order using explicit sequencing words such as first, then, after that, and finally, because discourse markers are visible signals of structure to an automated scorer. Keep speaking for the full expected duration; long silences cost more than a slightly imperfect sentence. If you lose a detail, do not stop to retrieve it, move forward and, if it returns, add it at the end.

The failure pattern is a candidate who tries to reproduce the original sentences verbatim, gets one word wrong, freezes, and delivers twelve seconds of audio for a forty second task. Coverage plus continuity beats precision here. Practise with short news clips or podcast segments, retell them immediately, and check your retelling against the source for missed facts.

RETELL SKELETON (fill these five slots while listening)
1. WHO   : the main person or group
2. WHERE : place and rough time
3. FIRST : the situation at the start
4. CHANGE: the event or problem that shifted it
5. END   : the outcome or resolution

SPOKEN TEMPLATE
"This passage was about <WHO> in <WHERE>.
 First, <FIRST>.
 Then, <CHANGE>.
 After that, <what followed>.
 Finally, <END>."

RULES
- Keep talking for the full duration, silence costs more than imperfection.
- Use first / then / after that / finally out loud, they signal structure.
- Paraphrase freely, do not chase the original wording.

Answer

Automated speech scoring systems generally evaluate a small set of measurable dimensions rather than judging you the way a human listener would. Broadly these are pronunciation, meaning how closely your phonemes match expected models, fluency, meaning speech rate, pause frequency and pause placement, prosody, meaning stress, rhythm and intonation contour, and for open tasks, lexical and grammatical range plus content coverage against the source. I am describing the general shape of such systems rather than the exact rubric of any specific vendor, because scoring weights are not published.

The practical implications are consistent regardless of vendor. Speak in complete sentences, because fragments break the prosody model. Place your pauses at clause boundaries rather than mid phrase, because random hesitation reads as disfluency while a pause after a comma reads as phrasing.

Keep volume and distance from the microphone constant, since level changes degrade recognition. Avoid filler sounds, because they count as disfluency, and replace them with a short silent pause. Do not over enunciate into a robotic staccato; unnaturally slow speech with equal stress on every word can score worse than natural speech, since prosody is a scored dimension. The honest summary is that clear, steady, complete sentences at a moderate pace optimise the scorer and also happen to be good communication.

Answer

A large share of low scores are environmental rather than linguistic. Test the microphone on the actual device you will use, at the same distance you will sit, and listen to the playback. Laptop built in microphones pick up fan noise and keyboard clicks; a basic wired headset with a boom microphone is usually a clear improvement over both a laptop microphone and cheap wireless earbuds, which compress audio heavily when the microphone is active.

Choose a small room with soft furnishings over a large tiled or empty room, because reverberation smears consonants. Close the window if there is traffic, silence phone notifications, and tell the household you are unavailable, since a single interruption can cost an entire task with no retry. On delivery, sit upright, because slouching compresses breath and volume drops off across a long sentence.

Keep water within reach, since a dry mouth changes articulation noticeably by the third task. Do not wear a mask or cover your mouth, and do not turn your head away mid sentence. What fails is treating this like a written test that happens to have audio. Give yourself fifteen minutes before the slot to do a full audio check and a short warm up by reading a paragraph aloud, so your first spoken words of the day are not the graded ones.

Answer

Name the language you have actually written the most code in, not the one you think sounds strongest. The moment you say Java, the panel will ask about OOPS concepts, the difference between an abstract class and an interface, collections including ArrayList versus LinkedList and HashMap internals at a basic level, exception handling with checked versus unchecked, string immutability and the string pool, and static versus instance members. If you say Python, expect list versus tuple versus set versus dictionary, mutability, list comprehensions, decorators and generators at a light level, shallow versus deep copy, and how exception handling works.

If you say C or C++, expect pointers, memory allocation, structures, and OOPS in the C++ case. The interviewer is probing depth in one area rather than breadth across many, so declaring three languages invites the panel to test the weakest. A weak answer is 'I know Java, Python, C, C++ and JavaScript', which almost always ends with a question you cannot answer in one of them.

A strong answer is 'Java is my primary language, I have used Python for scripting', which sets the scope of the conversation and still shows range. Prepare five to seven core answers deeply in your chosen language rather than twenty shallow ones across four.

Answer

The panel already knows the definitions; they are checking whether you have used the ideas. Anchor each pillar to your own project. Encapsulation is bundling data with the methods that operate on it and restricting direct access, for example a Payment class holding the amount and status as private fields with a markPaid method, so no other class can set status to paid without going through your validation.

Abstraction is exposing what a component does while hiding how, for example a NotificationService interface with a send method, where the caller does not know whether it is email or SMS underneath. Inheritance is reusing behaviour through a parent class, for example a base User with common fields and a Student subclass adding enrolment data, and the honest addition is that composition is usually preferred over deep inheritance in real code. Polymorphism is one interface with multiple behaviours, split into compile time overloading and runtime overriding, for example the same send call dispatching to EmailSender or SmsSender at runtime.

The interviewer is probing whether you can connect theory to code you wrote. A weak answer recites 'encapsulation is data hiding, abstraction is hiding implementation' with no example, and it is weak precisely because the two sound identical without one. Always follow the definition with the word 'for example'.

Answer

First normal form requires atomic column values and no repeating groups, so a column holding 'java, sql, aws' violates it and must become separate rows in a related table. Second normal form applies to composite keys and requires every non key column to depend on the whole key, not part of it, so in an order_items table keyed by order_id plus product_id, storing product_name is a violation because the name depends only on product_id. Third normal form removes transitive dependencies, where a non key column depends on another non key column, so storing city and pincode plus state in an employee table is a violation because state depends on pincode, not on the employee.

The follow up that separates candidates is denormalisation. You denormalise deliberately when read performance matters more than write simplicity: reporting tables, dashboards, analytics warehouses, or a frequently joined lookup that is effectively static. The cost is update anomalies, since the same fact now lives in two places and can diverge, so denormalised copies need a clear owner and a refresh path.

Interviewers also probe ACID here, so be ready with atomicity, consistency, isolation and durability plus one concrete example of each. A weak answer recites the normal forms without a single example table, which signals memorisation, and then cannot answer why anyone would ever denormalise.

Answer

This pair is close to a standard fresher SQL screen. The first query needs a join, an aggregate, a row level filter and a group level filter, and the discriminating detail is knowing that the active status filter belongs in WHERE while the headcount condition belongs in HAVING, because WHERE runs before grouping and HAVING runs after. Putting the status filter in HAVING either fails or forces the engine to aggregate rows it should never have read.

The second query is the classic second highest salary problem and there are three accepted answers. The subquery form takes the maximum salary strictly below the overall maximum, which handles duplicates correctly because MAX ignores ties. The window function form uses DENSE_RANK over the salary ordering, which is the right choice when you need the second highest per department rather than overall, and DENSE_RANK rather than ROW_NUMBER matters because tied salaries should share a rank.

The LIMIT with OFFSET form is the one to avoid without a DISTINCT, because with duplicate top salaries it returns the top value again. What the interviewer probes is whether you reason about duplicates unprompted. A weak answer writes ORDER BY salary DESC LIMIT 1 OFFSET 1 and, when asked what happens if two people earn the top salary, has no answer.

/* Departments with more than five active people, richest first */
SELECT d.dept_name,
       COUNT(e.emp_id)         AS headcount,
       ROUND(AVG(e.salary), 2) AS avg_salary
FROM   employees   e
JOIN   departments d ON d.dept_id = e.dept_id
WHERE  e.status = 'ACTIVE'        /* row filter, runs before grouping */
GROUP BY d.dept_name
HAVING COUNT(e.emp_id) > 5        /* group filter, runs after grouping */
ORDER BY avg_salary DESC;

/* Second highest salary overall, duplicate safe */
SELECT MAX(salary) AS second_highest
FROM   employees
WHERE  salary < (SELECT MAX(salary) FROM employees);

/* Second highest per department, ties share a rank */
SELECT dept_id, emp_id, salary
FROM (
  SELECT dept_id, emp_id, salary,
         DENSE_RANK() OVER (PARTITION BY dept_id
                            ORDER BY salary DESC) AS rnk
  FROM   employees
) ranked
WHERE rnk = 2;

/* Avoid: breaks when two people share the top salary */
/* SELECT salary FROM employees ORDER BY salary DESC LIMIT 1 OFFSET 1; */

Answer

This is the highest leverage question in the entire technical round because everything after it is drawn from your own answer. Structure it in four parts and keep the whole thing under two minutes before pausing for questions. One, the problem in one sentence framed as a user need, not a technology list.

Two, the architecture in three or four sentences: what the frontend was, what the backend was, what the database was, how data flowed. Three, your specific contribution stated in the first person singular, naming the modules you personally wrote, because in a four person project the panel wants your boundary, not the team's. Four, the hardest problem you hit and how you diagnosed it, which is the part that earns respect: a query that timed out and how you found it, a race condition on concurrent submissions, an authentication token that expired mid session, a deployment that worked locally and failed on the server.

Be ready for the follow ups: why that database, what would break at ten thousand users, how did you test it, what would you rebuild. What the interviewer is probing is whether you built it or watched it get built. A weak answer says 'we made a web application using React, Node and MongoDB with login and CRUD', which is a stack list, contains no ownership, and gives the panel nothing to dig into except your fundamentals.

Answer

The expectation is conversational fluency, not certification. Know the three service models with a real example each: infrastructure as a service where you rent virtual machines and manage the operating system, platform as a service where you deploy an application and the platform manages the runtime, and software as a service where you consume a finished product. Know deployment models: public, private, hybrid and community.

Know the core service categories in at least one provider among AWS, Azure and Google Cloud, meaning compute, object storage, managed relational and non relational databases, identity and access management, networking, and monitoring. Know the shared responsibility model, that the provider secures the cloud and the customer secures what they put in it, because it is a standard client conversation. Know why organisations migrate: capital expenditure moving to operational expenditure, elasticity for variable load, faster provisioning and managed resilience.

Understand vertical versus horizontal scaling, high availability versus disaster recovery, and roughly what a load balancer and an auto scaling group do. If you hold any associate level cloud certification, say so, because services organisations value them and often sponsor them after joining. A weak answer treats cloud as a synonym for 'a server somewhere' or lists service names without being able to say what any of them does, which is easy for a panel to catch with one follow up.

Answer

Less than a product company loop and more than nothing. The realistic surface is arrays and strings with confident manipulation, linked lists including reversal and cycle detection at a conceptual level, stacks and queues with their use cases, hash maps and where they beat arrays, binary trees with the three traversals, recursion with a clear base case, and the common sorting algorithms with their complexities. Big O notation should be automatic for the operations you name.

What is normally not required is dynamic programming beyond the simplest examples, advanced graph algorithms, segment trees or heavy competitive programming material. Where candidates lose marks is not depth but articulation: being unable to state the time complexity of a hash map lookup, or explaining a traversal by drawing nothing. Always take the whiteboard or the paper.

The interviewer is probing whether you can be trusted with logic in a delivery codebase, so a correct, clearly explained, straightforward solution outperforms a clever one you cannot justify. A weak answer is either a memorised optimal solution with no explanation, or 'I know the concepts but I need to look up the code', which reads as never having implemented anything. Implement each of the listed structures once by hand, and you will be comfortably above the bar.

Answer

Gen AI awareness questions have become common in services interviews because clients ask about it constantly, and panels want to know whether you have an informed opinion or a hype reflex. A strong answer has three parts. Where it helps: drafting and reviewing code, generating unit tests, summarising long requirement or defect threads, converting legacy documentation into structured formats, first draft translations, and retrieval augmented question answering over an internal knowledge base so support teams stop searching manually.

How you would build it responsibly: retrieval augmented generation over approved sources rather than fine tuning on client data by default, human review on anything customer facing, evaluation against a fixed test set before rollout, and logging of prompts and outputs for audit. Where you would refuse or insist on controls: sending client confidential data or personal information to a public model endpoint, letting a model take an irreversible action without human approval, and using generated output as fact in a regulated deliverable without verification. Naming hallucination, data residency, intellectual property in training data, and prompt injection as concrete risks shows you have thought past the demo.

A weak answer is either 'AI will replace everything' or 'we should not use it, it is unreliable'. Both are unusable to a delivery lead who needs a judgement call with guardrails.

Answer

This is a breadth probe. The panel is checking how many layers you can name coherently and where your knowledge actually stops. The chain: the browser parses the URL and checks its caches, then DNS resolution runs through the browser cache, the operating system cache, the configured resolver and the authoritative name servers, returning an IP address.

A TCP connection is opened through the three way handshake, and for HTTPS a TLS handshake follows where certificates are validated and session keys agreed. The browser sends an HTTP request with headers and cookies. That request may hit a content delivery network edge, a load balancer, a reverse proxy, then an application server, which may query a database or a cache and render or return data.

The response travels back with a status code and caching headers. The browser parses HTML into the DOM, CSS into the CSSOM, builds the render tree, computes layout, paints, and executes JavaScript, which may trigger further requests. What makes the answer strong is signposting the layers before diving in, so the interviewer can steer you to the one they care about, and being honest at your boundary: 'I can explain the TLS handshake at a high level, I have not implemented one'. A weak answer is a memorised paragraph delivered at speed with no pauses, which invites a follow up that immediately exposes the missing depth.

Answer

The managerial round exists to test client facing readiness, and this is its signature question. The panel is probing whether you escalate blindly, freeze, or make progress under uncertainty. The structure that works: clarify, document, decide, communicate.

First, attempt clarification with specific closed questions rather than an open 'please clarify', because a client who did not write a clear requirement will not write a clear clarification either. Ask 'should an expired coupon still apply to a partially shipped order, yes or no', with the two outcomes spelled out. Second, if clarification will take longer than your slack, state a documented assumption in writing, proceed on the lower risk interpretation, and flag it in the daily status so it is visible and reversible.

Third, protect the deadline by isolating the ambiguous piece behind a configurable switch or a clearly bounded module, so a reversal costs hours rather than a rewrite. Fourth, escalate early with options rather than late with a problem: present interpretation A, interpretation B, the cost of each, and your recommendation. A weak answer is 'I would ask my manager', full stop, which reads as no ownership, or 'I would build what I think is right', which reads as a candidate who will surprise a client. The strong answer shows movement plus traceability.

Answer

Accenture delivery teams routinely work with clients in North America, Europe, the Middle East, Japan and Australia, so time zone handling is a real operating skill and not a hypothetical. The answer the panel wants is that you treat the overlap window as the scarcest resource in the project. Use it only for things that genuinely need synchronous conversation: decisions, blockers, demos and disagreements.

Everything else, status, questions with a known answer, and code review comments, moves to asynchronous written channels. Batch your questions through the day rather than sending them one at a time, so the client answers a numbered list in one pass instead of drip feeding replies you will not see for eighteen hours. Write handoff notes at the end of your day covering what was completed, what is blocked and what you need answered before your next morning, because an unanswered blocker costs a full day, not an hour.

Be explicit about time zones in every message, writing 'Thursday 10:00 IST, which is Wednesday 23:30 Central' rather than 'tomorrow morning', since 'tomorrow' is ambiguous across the date line. If you have done this at an internship or in an open source project, say so concretely. A weak answer is 'I am flexible and can work any shift', which answers a different question and skips the working model entirely.

Answer

This is scored on ownership under pressure and on judgement about escalation, not on heroism. The panel wants to hear that you acknowledge fast, establish impact before touching anything, involve the right owner rather than working alone, and communicate at a fixed cadence so the client is never guessing. Acknowledging within minutes is disproportionately valuable, because most client anger during an incident comes from silence rather than from the defect itself.

Establishing impact means asking how many users are affected, whether there is a workaround, and whether data is being lost, since those answers determine whether this is a weekend war room or a Monday ticket. Involving the owner is not passing the buck; it is recognising that the fastest resolution runs through the person who knows the code, and your job in the meantime is triage, evidence collection and communication. The answer should end with the follow through: a root cause note and a preventive action, because managers are listening for whether you close loops.

A weak answer is either 'I would tell them it is not my module', which reads as a person a client will complain about, or 'I would fix it myself immediately', which reads as someone who will make an untested production change at 9 pm on a Friday. The sample below is the version to adapt.

SAMPLE ANSWER

"My first message would go out within a few minutes, just acknowledging that I have seen it and am checking, because silence is what actually escalates an incident. Then I would establish impact before touching anything: how many users are hit, is there a workaround, is any data being lost. With that, I would pull in the module owner through whatever the project's escalation path is, and while they are joining I would collect the evidence they will need, the error logs, the time the behaviour started, and what changed recently. I would keep the client updated on a fixed cadence, say every thirty minutes, even when the update is that we are still investigating. Once it is stable, I would write up the root cause and one preventive action, because if it happened on a Friday once it will happen again."

WHY THIS WORKS
- Acknowledges fast, which is what clients actually judge.
- Establishes severity before acting, so the response matches the problem.
- Involves the owner without abandoning responsibility for communication.
- Ends with prevention, which is what a delivery manager is listening for.

Answer

This question is scored on whether your reasons are specific to this kind of work or generic enough to apply to any employer. What the panel does not want is flattery about being a global leader, which every candidate says and which contains no information about you. What works is naming what services delivery actually offers that a single product team does not: exposure to multiple industries and client architectures in the first few years, a structured training and certification path, defined career bands with visible progression, and the chance to see how large enterprises actually run rather than one codebase.

Then connect that to something concrete about you, ideally your project or internship. Then acknowledge the trade off honestly, because a candidate who has thought about it is more convincing than one who claims everything is perfect: services work means the client sets priorities, you may work on maintenance before greenfield, and the technology is chosen by the engagement. Saying you understand that and still want it is more persuasive than pretending the trade off does not exist. A weak answer is 'Accenture is a reputed company with good growth opportunities', which is interchangeable across every employer and tells the panel you would accept any offer, which they already assume and would prefer not to have confirmed.

SAMPLE ANSWER

"Two reasons, one about the work and one about me. On the work, in a services role I get to see several industries and several architectures in my first few years instead of one codebase, and that breadth is exactly what I want early on before I specialise. The structured training and certification path matters to me too, because I learn faster when there is a defined curriculum rather than only self study. On me, my final year project was a claims tracking tool I built for a small insurance office, and the part I enjoyed most was not the code, it was sitting with the clerk who would use it and working out what she actually needed. That is closer to client facing delivery than to building for an anonymous user base. I also know the trade off, that in services the client sets the priorities and the stack, and I am comfortable with that at this stage of my career."

WHY THIS WORKS
- Gives a reason specific to services work, not generic praise.
- Anchors it in a real project, so it cannot be recycled by another candidate.
- Names the trade off honestly, which reads as a considered choice.

Answer

HR is checking deployability, meaning whether you can be staffed on any project without a constraint that has to be managed. Delivery centres are spread across Bengaluru, Hyderabad, Chennai, Pune, Mumbai, Gurugram, Kolkata and Coimbatore, and client coverage across North American and European time zones means some projects run shifted hours. The answer that scores well is an unqualified yes with a short, non negotiating context line, not a yes wrapped in conditions.

If you genuinely have a constraint, for example a medical dependency at home, state it once, clearly, and immediately state what you can do, rather than hiding it and creating a problem at allocation time. Do not negotiate location in the HR round; it is the wrong stage and the wrong person, since staffing decisions happen at allocation, not at offer. Do not say 'I prefer Bengaluru but I can adjust', because the word 'prefer' is what gets recorded.

What a weak answer sounds like is a conditional yes with a list, 'yes, but not night shifts and preferably South India', which converts you from a flexible candidate into a constrained one in a single sentence. If you are truly location bound, apply to drives for that location rather than trying to negotiate a general offer.

SAMPLE ANSWER

"Yes to both. I am open to any of the delivery locations, and I have already discussed relocation at home so it is a decided thing, not something I would need to reconsider after the offer. On shifts, I understand that client coverage across US and European time zones means the working window shifts by project, and I am fine with that. During my final semester I was doing lab work till late and starting client calls for my internship at eight in the morning, so I have managed an irregular schedule before and I know what it takes to stay consistent through it."

WHY THIS WORKS
- Unqualified yes on both, which is what deployability means.
- Shows the family conversation already happened, removing the usual risk.
- Backs flexibility with a lived example rather than a promise.
- Never uses the word prefer, which is what gets written down.

Answer

For laterals this question decides your start date and is a real commercial input, since a long notice can cost a candidate a role when a client position needs filling quickly. Answer with three facts in order: the contractual notice, whether buyout is permitted by your current employer, and the earliest realistic date you can join. Be precise, because 'around two to three months' reads as uncertainty about your own contract.

If buyout is possible, say so and say who pays under what conditions, and ask rather than assume, since reimbursement policies vary by role, band and drive and are not guaranteed. On the counter offer question, HR is testing offer stability, because a candidate who accepts and then withdraws costs the recruiter a completed pipeline. Answer it directly: say you have already thought about it, state the reason you are moving in terms the counter offer cannot fix, for example scope, domain exposure or technology direction rather than money, and then say plainly that you will not be accepting one.

Do not say 'it depends on the number', which is honest but tells HR you are using this offer as leverage. For freshers, the equivalent is 'do you have other offers', and the safe answer is truthful about having applied elsewhere while being clear on your interest here. A weak answer is a vague date and hedged commitment; both invite a slower offer.

SAMPLE ANSWER

"My contractual notice is 60 days from the date I resign. Buyout is allowed at my current company for up to 30 of those days, so the earliest realistic joining date is about 30 to 35 days from an offer, and I would confirm the exact figure with my HR the same week. On a counter offer, I have already thought that through. The reason I am moving is that I want enterprise scale delivery exposure and client facing work, and my current employer cannot offer that regardless of what they revise on salary. So no, I would not accept a counter offer, and I would give you a firm resignation date rather than keeping you waiting."

WHY THIS WORKS
- Gives exact numbers, so recruiting can plan an onboarding date.
- Raises buyout as a question rather than assuming reimbursement.
- Frames the move around scope, which a counter offer cannot match.
- Commits explicitly, which is what the counter offer question is testing.

Answer

Background verification for an IT services role in India typically covers identity and address, education, meaning degree and marksheets verified with the institution or a verification agency, previous employment, meaning dates, designation and reason for leaving verified with the prior employer, and often a criminal record check. For laterals, relieving letters and payslips are commonly requested. Verification is normally conducted by a third party agency and can run in parallel with or after joining, depending on the drive.

What matters for you is disclosure timing. Discrepancies that would have been survivable if declared up front become serious when the agency finds them: overstated designation, dates padded to hide a gap, an internship listed as full time employment, or an unresolved backlog described as cleared. Declare gaps, arrears, an ongoing case, or a company that has shut down and cannot verify, and declare them before the check begins rather than after a flag.

Keep a document folder ready with all semester marksheets, degree or provisional certificate, government identity, PAN, passport size photographs, and for laterals the offer letter, relieving letter and recent payslips from every employer. A weak approach is assuming small inaccuracies will not be caught. Verification agencies are checking exactly those fields, and an integrity flag is materially harder to recover from than the original gap would have been.

Answer

They are listening for three things: whether you can structure a spoken answer, whether your story leads logically to this role, and whether you sound like you will be easy to work with. They are not listening for your date of birth, your father's occupation or your hobbies unless asked. Use a four part structure of about ninety seconds: who you are academically right now, one thing you have built or done with a concrete detail, one strength evidenced rather than asserted, and why this role follows from the previous three sentences.

Concrete details are what make the answer memorable, so 'a Django attendance system used by 300 students in my department' beats 'various projects'. Evidence beats adjectives, so 'I was the one who kept our four person project on schedule by running a fifteen minute check in every morning' beats 'I have good leadership skills'. Then stop talking.

Rambling past two minutes is the most common failure here, because the panel stops listening and you have given away control of the conversation. What a weak answer sounds like is a chronological recitation from birth, school and marks, delivered flatly, with no link to the job. Rehearse it aloud until it is natural rather than memorised, since a visibly recited answer creates exactly the impression you were trying to avoid.

SAMPLE ANSWER

"I am a final year Information Technology student at a college in Pune, graduating this year with around 78 percent. The project I am most proud of is an attendance and internal marks portal I built with Django and PostgreSQL that my department has actually been using, about 300 students and 14 faculty members. I wrote the backend and the reporting module. The thing I would call my strength is follow through: on that project I ran a fifteen minute stand up every morning with my three teammates, which is why we shipped a working version two weeks before the deadline while most groups were still integrating. I am applying to Accenture because I want client facing delivery work across different industries early in my career rather than one product, and I want the structured training path that comes with it."

WHY THIS WORKS
- Ninety seconds, four parts, ends on why this role.
- Uses specific numbers, which make it impossible to confuse with another candidate.
- Evidences the strength with a behaviour instead of asserting an adjective.
- Leaves obvious hooks the panel can follow up on, so you steer the round.

Answer

Accenture has run more than one entry band for freshers rather than a single uniform fresher offer, and candidates commonly discuss two: the Associate Software Engineer, usually referred to as ASE, which is the standard entry role, and a higher entry band commonly referred to as Advanced Application Engineer, offered to candidates who clear an additional or harder assessment stage. As of 2026, the pattern reported by candidates is that the standard entry role sits in roughly the ₹4.5 to ₹5 LPA range while the higher entry band sits materially above it, often quoted in the ₹6.5 to ₹9 LPA range, and specialised or niche roles can go higher. I would treat those bands as indicative only.

Actual figures vary by year, by drive, by degree and by whether the offer is campus or off campus, and Accenture does not commit to a public fresher pay scale. What matters tactically is that the higher band is normally gated behind extra assessment performance, so if a drive offers an optional additional round, attempt it, since the downside is usually just time. If you are asked in an interview which role you want, say the higher band and give a reason tied to the work, not to the money. A weak answer treats the two as identical or quotes a forum number as if it were official policy.

Answer

Large services firms hire against forecast demand, so the gap between an offer letter and a date of joining can range from a few weeks to several months, and candidates have at times seen extended waits during slower demand cycles. Treat the wait as a managed period rather than dead time. Practically, first, stay in contact through the official channel only, meaning the recruitment contact or candidate portal, and check in at a reasonable cadence such as once every three or four weeks rather than weekly, because pressure does not advance an allocation.

Second, complete every onboarding formality the moment it is requested, since incomplete documentation is a genuine cause of delay that is entirely within your control. Third, keep preparing: many firms provide pre joining learning material or certification paths, and completing them is visible and useful. Fourth, and this is the important one, do not stop your job search until you have a confirmed date of joining in writing.

Holding a single undated offer as your only plan is the mistake that costs candidates a year. Fifth, if you accept another role in the meantime, decline formally and politely rather than disappearing, because services HR teams keep records and you may reapply later. Deferred joining is a demand cycle outcome, not a judgement on you, so do not read it as a signal about your performance.

Answer

The typical shape reported by new joiners at large Indian services firms is an induction covering policies, security and compliance, followed by a structured technical training programme, then an assessment at the end of training, then allocation to a project. Training content is usually assigned by track, for example application development, testing, data, cloud or infrastructure, and performance in training can influence which project you are allocated to, which is why it matters more than it appears. Probation and confirmation timelines vary by employment terms, so read your own offer letter rather than relying on what a batchmate says.

Three things are worth optimising for. First, training assessment performance, because early allocation to a project on a technology you want is far easier than transferring later. Second, visible reliability, meaning you respond, you finish what you said you would, and you flag blockers early, since in a batch of hundreds, being dependable is what gets you noticed by a delivery lead.

Third, certifications, because services organisations value and often sponsor them, and they are a concrete lever on your first appraisal. What new joiners get wrong is treating training as a formality and coasting through it, then being surprised by a maintenance allocation on a technology they did not want.