REAL CANDIDATE REPORTS — 2026 SITTINGS

Every one of 2,133 messages read · a 387-member TCS NQT 2027 group · 15 May – 2 Aug 2026

This is the best evidence in the project. Not prep-site marketing, not a historical corpus — students who sat the exam in June and July 2026, posting within hours of walking out.

It corrects four things in my earlier notes and adds a large amount of exam mechanics no prep site publishes.

⚠️ CORRECTION 1 — YOU CAN REVISIT QUESTIONS (except in Verbal)

I previously told you navigation was locked everywhere. That is wrong.

"Can we navigate through questions? Can we revisit them?" → "Yes except verbal. And not navigate between sections." — 2 July "Will rough sheets be provided? Can we revisit questions after attempting or after leaving it unanswered?" → "Yes rough sheets are provided. And you can revisit. But not the verbal section." — 20 July "Is the timer section-wise, or 1-minute per question? Can we go back?" → "In aptitude and reasoning you don't have a time limit for individual questions, whereas in Verbal Ability you will find the timer." — 22 July

The accurate picture

What changes for you: in aptitude and reasoning you can skip a hard question and come back. My "commit and move, there is no review pass" advice was too strong for those sections — it is correct only for Verbal.

Strategy: in Numerical, do a fast first pass taking the easy marks, then return to the hard ones with whatever time is left. In Verbal, there is no return — answer within 25 seconds or lose it.

⚠️ CORRECTION 2 — SENTENCE COMPLETION HAS NO OPTIONS

"There were 25 sentence completion questions, and there were no options; we had to enter the answer directly in the input box." — 9 June "No option available, you have only 25 seconds, after that you can't attempt — it will lock." — 9 June "There is no options bro, we have to type the word and each question stays for only 25 secs." — 13 June

From a trainer's slide sourced to an exam document:

"Type ONE word that best fits the sentence. Only one word accepted. Click 'Next' when done." — 20 Qs · 25 sec each

This voids option-elimination for verbal

"Eliminate options", "solve the easier blank first", "predict then find the closest option" — all assume four choices. Cross them out.

Instead: vocabulary must be active recall (produce the word, don't recognise it); grammar becomes decisive; spelling counts — you are typing.

One candidate on difficulty: "Verbal bhi easy tha, fill in the blanks are super easy." — 19 July And on content: "was verbal based on grammar or just vocabulary?" → "Just vocabulary" — 25 July

⚠️ CORRECTION 3 — SIMPLE & COMPOUND INTEREST IS A TOP TOPIC

My corpus put SI at 23 questions → Tier 3, skim it. Five independent reports say otherwise:

"Simple Interest (Approx. 6 Questions)" — 7 July, Slot 1 "Simple Interest (More than 5 questions)" — 8 July drive, Slot 2 "Simple & Compound Interest" listed first — 8 July, Slot 1 and Slot 2 "Numerical me kaunse topic se aaya tha?" → "SI CI, mensuration" — 19 July "Percentage aur SI & CI se hi maximum questions the" — 24 July

→ Promoted to MUST MASTER, moved to Day 5. It is pure formula and takes about an hour.

⚠️ CORRECTION 4 — CODING INCLUDES TREES, HEAPS, STACKS AND GRAPHS

My corpus deprioritised these. The 2026 reports contradict it.

"In the first shift they asked Binary search tree for the easy question itself. So be well prepared in arrays and basic idea about trees." — 20 July "Trees was asked in one of the shifts in June. But if you are lucky enough it will not be asked." — 18 July "Which major DSA topics are asked?" → "Array, string, linked list, stack, queue, hashmap, hashset" — 21 July "Just focus on arrays, strings, number system, sorting… if you want PRIME role then you should also focus on graphs, trees." — 28 June

📋 EVERY CODING QUESTION REPORTED, 2026

Add to your prep: binary tree basics + min depth + BST · heap/PriorityQueue · BFS on a grid · merge sort and quick sort by hand · binary-search variants (first occurrence, peak, k-th) · matrix transformations · Roman numerals · stack implementation.

🔴 THE SINGLE BEST PIECE OF ADVICE IN THE ENTIRE CHAT

"Guys, coding section — focus on input handling. You need to handle exceptions while getting errors. In hard-level coding problems the logic will be easier — the twist is kept in the input getting. You will face compiler issues but that's not real…" — 19 July

This validates §2 of the coding file completely. And here is the proof, from the same day:

"Smallest missing positive element in the array. Constraint: the array may contain float values; if it contains even one float, print 'Invalid Input'. Input: -29,-24,4,2,1,9.9,24,43,5,6,7 — no n was given. You were just given the string. Now take this as String input, split it, check for float values… TBH the question was very easy. But the vague problem description, improper constraints and testcases, and the input handling make it hard."

"Yes it worked for me. I took the complete string as input using sc.nextLine() and then split it using (,)" — exactly the universal-reader approach.

"Input ka format kya tha? In my shift the input was in String. Many people were trying to take input in nextInt() and they were not able to solve the problem. The input was comma separated 1,2,3,4,5 — people thought they were getting integer input one by one." — 21 July

🖥️ THE COMPILER — what actually goes wrong

⭐ The Java class name

"It will be mentioned in the question at the end — name the class as MyClass. If nothing is mentioned, go with Main." — 8 July "If you are coding in Java always use the class name MyClass as mentioned in the instructions. Many people were unaware and submitted with a different class name, which may have caused them to lose marks." — 9 June

Read the bottom of the question statement for the required class name.

⭐ The trailing-space bug — it failed test cases en masse

On 28 June many candidates had correct Fibonacci logic and still failed test case 2. The group diagnosed it:

"You are printing a space with the last character as well. Output should be 5 6 11 17 28 45 73 but you are printing 5 6 11 … 73 (trailing space)."

System.out.print(x + " ") in a loop leaves a trailing space. Build with a delimiter or trim before printing.

⭐ Arrays failing where loops work

"I did the Fibonacci by taking an array but got array index out of bounds… it's correct code, works here but not in the TCS compiler." "You have to do the code without array."

The group's own diagnosis was correct: hardcoding arr[0]=5; arr[1]=6; throws when n = 1. Handle n = 1 and n = 2 explicitly — or avoid the array and use two rolling variables (which is what my Fibonacci solution does).

Other platform facts

• You can compile and test multiple times — "bas last time jo compile karenge wo consider hoga" (the last compile counts)
• Write everything from scratch — "you even have to write main class, import modules, psvm, everything"; the binary-tree question came with "a completely empty file — no boilerplate, no I/O handling, no function templates"
• C++: #include<bits/stdc++.h> gives an error — import the specific headers
• Java: import java.util.*; worked for one candidate; another claimed it didn't. Import specifically if in doubt.
• Built-in functions are allowed — one candidate used Arrays.sort() successfully
• Compilation can take ~10 seconds per run; on some days it failed outright and one shift's exam was cancelled and re-held

🖥️ EXAM MECHANICS

Section order is FIXED

Numerical → Reasoning → Advance Aptitude → Verbal → Coding (easy) → Coding (medium)

"No you cannot change order." — 21 July

Numerical is first — the hardest section, when you are most nervous. Reasoning and Verbal (the easy ones) come after. Do not let a bad Numerical section poison the rest.

✅ Calculator and rough sheets ARE provided

"Is a screen calculator provided?" → "Yes" · "Will we get rough sheets?" → "Yes" "Calculator use kar sakte the. Computer pe allowed hota hai." — 19 July But: "that calc is just a waste of time — try to get small calculations done by yourself."

Practise method selection, not long multiplication. And the "write the alphabet table first" tactic works, because rough sheets exist.

⚠️ Non-MCQ answers: whole numbers, no symbols

From official instruction text a candidate photographed:

"This assessment has a mix of MCQ type with options and non-MCQ type. For non-MCQ, enter the answer using the on-screen keyboard… type answers in whole numbers without using any symbols (e.g. %, kms, ms, ₹)."

20% → type 20. ₹8,640 → type 8640. If you land on a fraction, re-check your method. (For the typed verbal answers a candidate confirmed you get a physical keyboard.)

The verbal breakdown (trainer slide, sourced to an exam doc)

Candidates reported 3, 4 and 5 passages across slots, and passages of "3–4 lines — you have to read very fast." You may write in your own words.

Real email topics reported: request an extension after missing a project deadline · request leave as a Sales Executive · a restaurant review as a customer · as a manager, tell employees to follow office rules (personal printing, web surfing) · environment/networking themes.

⏱️ TIMING — brutal, and worse than any prep site admits

"Out of 20, if you solve the first 5 questions there is no time remaining for the others." — 28 June "25 min mein toh questions bhi na padh pao" [in 25 minutes you can't even read the questions] · "Sabne tukke hi lagaye the" [everyone just guessed] "Almost 4 to 5 questions take 15 min." · "for 4 questions it took me 10 min" "I was able to solve only 2 questions in Numerical Aptitude and the rest 18 were guesswork. Logical and Verbal were easy." — 19 July "Numerical bahut hard tha — CAT level and bahut lengthy." — 19 July "Time is not sufficient for all sections except coding and email writing." "In aptitude, to solve all, you have to be at least like Shakuntala Devi."

What to take from this

1. You are not expected to finish Numerical. Nobody does.
2. Guessing the rest is correct strategy — there is no negative marking.
3. Reasoning and Verbal are where you bank marks. "Logical reasoning was very easy, you can score above 15 easily."
4. Because you can revisit within a section, sweep the easy ones first, then return.

📊 SELECTION RATES — real numbers

Cutoffs are shift-dependent: "Depends on how tough your shift is and how your batch mates performed — each shift has different cutoffs." On coding: "Solving 2 questions is enough to reach cutoff" (broadly agreed, not authoritative). Test cases reported as 5 and 7 depending on the question. Role guidance: "For digital, both apti + coding are required." · "If you have not started DSA, focus on arrays and strings and most importantly TCS PYQs — by this you can get up to the digital role."

Results arrive 2–5 days after the exam, and interviews sometimes the next day — one candidate had "less than 24 hrs to prepare."

🎤 THE INTERVIEW — one detailed account

"SQL, basic coding, pillars of OOPs with example, CN basic, OS basic, DBMS basic — no need to go in depth. No intro, no project-related questions were asked. So be prepared for the above things (only basic)." — 27 July "Just search top 15 interview questions and do that only… Jo maine upar bataya hai usse bahar se kuch nahi puchega, guarantee."

One reported question: "#include <stdio.h> — what is #?" Another common one: swap two variables without a temp (a=a^b; b=a^b; a=a^b; or the +/− method).

Bring original documents — "guys sometimes raise objections on duplicates, so it's better to have your originals."

Honest limits on this source

• It is a student group, not TCS. Reports are from memory, sometimes hours later, occasionally second-hand.
• Slot variation is enormous. One shift got a BST as the easy question; another got peak-element and merge sort. "It depends on your luck also."
• Some claims conflict and I have shown the ranges rather than picking: 3 vs 4 vs 5 passages · 5 vs 7 test cases · 20 vs 25 sentence completions · whether import java.util.* works.
• Nobody among 387 people could state the sectional cutoffs. Treat any specific cutoff figure — anywhere — with suspicion.
• Roughly 80% of the 2,133 messages were registration, eligibility and admin chatter with no exam content; the findings above are concentrated in the ~150 messages posted by people who had just sat the paper.
• Reports cover 2026 sittings of the 2027-batch drive — the closest available evidence to your exam, but not a guarantee.