Data sufficiency

In short

Data sufficiency (DS) asks you to decide whether given information is enough to answer a question - not to produce the answer itself. A question stem is followed by two (sometimes three) statements, and you must judge each statement alone and then in combination against a fixed answer key. It is a standing question type in MBA entrances (CAT, XAT, NMAT, SNAP, CMAT, MAT), bank and insurance exams (IBPS/SBI PO and Clerk, RBI Grade B, NABARD), SSC and railway reasoning papers, and campus aptitude platforms (AMCAT, CoCubes, eLitmus, TCS NQT, Infosys, Wipro, Capgemini, Accenture). It is also one of the few aptitude formats that directly mirrors a real workplace skill - deciding, before you start work, whether the data in hand can support a decision.

Explained

Anatomy of a data sufficiency question

• Stem (the question): either a *value* question ("What is the speed of the train?") or a *yes/no* question ("Is n an odd number?").
• Statements: numbered (1) and (2), or I, II, III. They are always true - never dispute them, never check whether they are realistic.
• Task: mark a code from the key. You are graded on the sufficiency judgement, not on the numerical answer.
• Statements never contradict each other or the stem. If statement (1) gives speed 20 m/s and statement (2) gives 25 m/s, you have made an arithmetic slip - go back.
• A statement that merely restates the stem or another statement is redundant and adds no sufficiency.

The two sufficiency tests

• Value question: a statement (or combination) is sufficient only if it narrows the answer to exactly one value. Two possible values = insufficient, however close they are.
• Yes/no question: sufficient if every case allowed by the statement gives the same answer. A definite NO is just as sufficient as a definite YES.
• Proving insufficiency: find two permitted cases that give different answers (two values, or one yes and one no). One counterexample pair settles it; endless algebra does not.
• Do not compute the final answer. Once you know a unique answer exists, stop. Solving wastes the 45-60 seconds a campus test allows per question.

The standard five-option key and the AD/BCE grid

The classic (GMAT/CAT/XAT) key:

• A - statement (1) alone is sufficient, (2) alone is not.
• B - statement (2) alone is sufficient, (1) alone is not.
• C - both together are sufficient, but neither alone.
• D - each statement alone is sufficient.
• E - both together are still not sufficient.

Method - the AD/BCE grid:

1. Read the stem. Identify the unknown and what would pin it down *before* reading the statements.
2. Test statement (1) alone. If sufficient, the answer is A or D; if not, it is B, C or E.
3. Test statement (2) alone - forgetting statement (1) completely. This is the single commonest error.
4. If both alone are sufficient → D; only one → A or B.
5. If neither alone works, combine them. Sufficient → C; still not → E.

Indian exam variants of the key - read it every time

• Bank two-statement format (IBPS/SBI): options are typically (a) I alone (b) II alone (c) either I or II (d) neither I nor II (e) both I and II together. Here "either I or II" = the classic D, and the order of options differs from the GMAT key - never answer from muscle memory.
• Three-statement format: "Which of the statements is/are sufficient?" with options such as *only I and II*, *only II and III*, *any two of the three*, *all three*, *even all three together are not sufficient*.
• Minimum-set principle: in three-statement sets you must report the *smallest* sufficient set. If I + II work, do not mark "all three".
• Redundancy example - who is tallest among P, Q, R? I: P > Q; II: Q > R; III: P > R. I + II → P > Q > R, sufficient. I + III → P beats both, sufficient. II + III → tallest is P or Q, insufficient. Correct answer: "I and II, or I and III".
• SSC/railway style: usually two statements with a four-option key; some papers ask only "is the data sufficient - yes or no".

Sufficiency logic for quantitative stems

• Degrees of freedom: u unknowns need u independent, consistent equations. x + y = 10 and 2x + 2y = 20 are one equation, not two → insufficient.
• Ratios carry no scale: a : b = 5 : 4 never gives a. You need one absolute quantity (a, b, a + b or a − b).
• Averages: mean = sum ÷ count. Knowing the sum and the count is sufficient; knowing the highest, lowest, median or range is not.
• Non-linear traps: x² = 36 → x = ±6 (insufficient) unless a statement fixes the sign, or the stem says "positive integer". |x − 3| = 5 → x = 8 or −2.
• Integer and positivity constraints in the stem are free data - "n is a two-digit prime" can turn an otherwise weak statement into a sufficient one.
• Percentages: final = initial × (1 + r). Any two of the three fix the third; a percentage change alone never gives an absolute value.
• Geometry: similar triangles / angles alone give ratios, not lengths; "area of a rectangle" needs l × b, which can come from (l + b) with (l² + b²) or with l − b, without ever finding l and b separately.
• Work, time, speed: speed needs distance and time, or two consistent crossing conditions; combined rates need individual rates or their sum.

Sufficiency logic for reasoning stems

• Blood relations: a relation is determined only when gender and generation are both fixed. "A is B's child" plus "A is C's brother" still leaves C's gender open.
• Directions: two perpendicular displacement components fix the position completely - "B is 3 m east and 4 m north of A" is sufficient for the direction (north-east), for the exact bearing (arctan ¾ ≈ 36.87° east of north) and for the distance (5 m). Sufficiency does not care that a computation is needed; one missing component, or a direction with no length, is what makes data insufficient.
• Comparisons/rankings: to name the top element among n items, every other item must be beaten by one candidate, directly or transitively - a full chain linking all items is not required (P > Q and P > R names P without comparing Q and R), but a single unlinked item destroys sufficiency.
• Seating and puzzles: a statement is sufficient if it kills all but one arrangement. Count arrangements, do not draw them all - "two mirror-image seatings remain" = insufficient for left/right questions, sufficient for "who is opposite whom".
• Coding-decoding: one coded word usually permits several rules; sufficiency needs enough letter-position overlap to fix the rule uniquely.
• Dates/days: a day of the week is fixed by a reference date plus a count of days; "the 1st was a Monday" plus the month length is sufficient for any date in that month.

Advantages, limitations and how examiners use it

• Advantages: short to read, tests reasoning rather than calculation speed, hard to crack by back-solving options, and no partial calculation is rewarded - so it separates careful candidates sharply.
• Limitations: a single misread of the option key loses a correct deduction; negative marking (0.25 mark in IBPS/SBI and XAT, 1/3 of a mark on CAT MCQs, none in NMAT - verify the current pattern in the notification) punishes guessing on C vs E, the pair most often confused.
• Examiner's favourite traps: the "obviously needed but actually redundant" second statement (answer A, not C), the hidden second root, and the yes/no question that looks like a value question.
• Standards note: there is no statutory syllabus for DS; the format follows each exam's notification, and option wordings have changed between IBPS cycles - always read the instruction block in the paper.

Key points

• Data sufficiency - judge whether information is enough; never produce the answer.
• Statements are true - you test their power, not their truth.
• Unique-value rule - a value question needs exactly one possible value; two means insufficient.
• Yes/no sufficiency - a consistent NO is sufficient; mixed outcomes are insufficient.
• Counterexample method - two permitted cases with different answers prove insufficiency instantly.
• AD/BCE grid - test (1) first; sufficient → A or D, insufficient → B, C or E.
• Independence of statements - evaluate (2) with amnesia about (1); carrying over is the top error.
• No contradiction rule - statements that disagree signal your own mistake.
• Redundant statement - restates known data; cannot create sufficiency.
• Degrees of freedom - u unknowns need u independent, consistent equations.
• Ratio needs scale - ratios plus one absolute value give the quantity.
• Average needs sum and count - extremes, median and range do not suffice.
• Non-linear multiplicity - squares, moduli and quadratics commonly give two answers.
• Stem constraints count - "positive", "integer", "prime", "two-digit" are usable data.
• Minimum set - in three-statement questions report the smallest sufficient combination.
• Key order varies - GMAT/CAT key and the IBPS key are differently ordered; read both.
• Stop early - sufficiency decided means question over; do not solve.
• C vs E discipline - combine only after both statements have failed alone.

Formulas and examples

Example 1 - value question, answer A. *What is the average mark of 5 students in a test?* (1) The total of their marks is 310. (2) The highest mark is 80 and the lowest is 50.

• (1): mean = 310 ÷ 5 = 62 - a single value → sufficient.
• (2): totals from 50 + 50 + 50 + 50 + 80 = 280 to 80 + 80 + 80 + 80 + 50 = 370 → many means → insufficient.
• Answer: A. (The temptation is C - statement (2) is real information, but it is not needed.)

Example 2 - value question, answer C. *What is the area of a rectangle?* (1) Its perimeter is 40 cm. (2) Its diagonal is √200 cm.

• (1): l + b = 20 → 19 × 1 = 19 cm² or 10 × 10 = 100 cm² → insufficient.
• (2): l² + b² = 200 → (10, 10) gives 100 cm², (√199, 1) gives ≈14.1 cm² → insufficient.
• Together: (l + b)² = 400 ⇒ 200 + 2lb = 400 ⇒ lb = 100 cm² → sufficient.
• Answer: C. Note you never needed l and b individually (they happen to be 10 and 10).

Example 3 - value question, answer D. *A train is 180 m long. What is its speed?* (1) It crosses an electric pole in 9 seconds. (2) It crosses a 270 m long platform in 22.5 seconds.

• (1): v = 180 ÷ 9 = 20 m/s = 20 × 18/5 = 72 km/h → sufficient.
• (2): v = (180 + 270) ÷ 22.5 = 450 ÷ 22.5 = 20 m/s = 72 km/h → sufficient.
• Answer: D. Both give 72 km/h - as they must, because statements cannot contradict.

Example 4 - yes/no question, answer E. *Is the positive integer n odd?* (1) n is divisible by 3. (2) n is divisible by 9.

• (1): n = 3 (odd) or n = 6 (even) → insufficient.
• (2): n = 9 (odd) or n = 18 (even) → insufficient.
• Together: n must be a multiple of 9 → 9 (odd) or 18 (even) → still insufficient.
• Answer: E. Statement (2) already contains (1), so combining adds nothing - a classic redundancy trap.

Example 5 - three-statement bank format. *Who scored the highest among P, Q, R and S?* I. P scored more than Q. II. R scored more than S. III. Q scored more than R.

• I + III: P > Q > R, but S is unplaced and could top the list → insufficient.
• II + III: Q > R > S, P unplaced → insufficient.
• I + II: P > Q and R > S, two separate chains → insufficient.
• I + II + III: P > Q > R > S → P is highest → sufficient.
• Answer: all three statements are required (and the minimum set is all three).

Where it is used

• Campus aptitude tests: AMCAT logical ability, CoCubes, eLitmus, TCS NQT, Infosys, Wipro NLTH, Capgemini, Accenture and Cognizant reasoning sections carry DS or DS-style sufficiency items.
• Government and banking recruitment: IBPS and SBI PO/Clerk reasoning, RBI Grade B, NABARD, LIC AAO, SSC CGL and RRB NTPC reasoning papers.
• MBA entrances and B-school coursework: CAT (QA/DILR), XAT, NMAT, SNAP, CMAT, MAT, MICAT; carried forward into case-interview "what data would you need?" questions.
• Business analyst and consulting work: scoping a requirement - deciding whether the fields available in a dataset can answer the business question before any model is built (common at TCS, Infosys, Deloitte, EY, KPMG analytics teams).
• Credit, audit and compliance roles: a bank credit officer judging whether submitted documents suffice to sanction a loan, or an auditor deciding whether available evidence supports an opinion, is performing data sufficiency.
• Data engineering and MIS: validating that a report's inputs are complete and non-redundant before publishing - the degrees-of-freedom idea applied to reconciliations.

Interview questions

• Q: What exactly does a data sufficiency question test? A: Whether the given information is enough to determine a unique answer - not the answer itself. You judge each statement alone, then together, against a fixed key.
• Q: When is a statement sufficient for a yes/no question? A: When every case allowed by the statement yields the same answer; a definite "no" is as sufficient as a definite "yes".
• Q: What is the fastest way to prove insufficiency? A: Produce two cases consistent with the statement that give different answers - one counterexample pair ends the question.
• Q: Why must statement (2) be evaluated without statement (1)? A: Because options A, B and D depend on standalone sufficiency; carrying information across makes (2) look stronger than it is, turning a true A into D or a true C into B.
• Q: Can the two statements contradict each other? A: No. In a well-set question both are true simultaneously, so a contradiction means you have miscalculated.
• Q: "What is x?" with (1) x² = 25 and (2) x > 0. Answer? A: C - (1) gives ±5, (2) gives infinitely many values, together x = 5.
• Q: If two linear equations in two unknowns are given, is the data always sufficient? A: No. If the equations are dependent (one a multiple of the other) they fix only a line, not a point.
• Q: A question asks for the average of 7 numbers and a statement gives the median and range. Sufficient? A: No. The average needs the sum; median and range are consistent with many different sums.
• Q: How do you handle a three-statement bank question? A: Test every pair before the triple and report the smallest sufficient combination; several different pairs may each be sufficient.
• Q: Is "the ratio of A's and B's salary is 3 : 2" ever sufficient for A's salary? A: Only if combined with an absolute figure - A + B, A − B, or either salary. A ratio has no scale.
• Q: Why is C the most frequently wrong answer students mark? A: Both statements usually look relevant, so candidates combine them without first checking whether one alone already pins the answer down (A, B or D) or whether even both leave two cases (E).
• Q: How does this skill show up on the job? A: Before building a report or model you ask what minimum data determines the decision; it prevents collecting redundant data and stops you from promising a conclusion the data cannot support.

Common mistakes

• Carrying statement (1) into statement (2) - the single biggest source of wrong D and B answers; consciously reset after step 2.
• Actually solving the problem - computing the full answer burns time that DS does not reward, and tempts you to overlook the second root.
• Missing the second case - forgetting negative values, zero, fractions or the ± from squares and moduli, and so calling an insufficient statement sufficient.
• Ignoring constraints in the stem - "positive integer", "prime", "two-digit", "distinct" are free data that often make a weak statement sufficient.
• Marking C reflexively when one statement alone was enough (answer A, B or D), or when even the combination leaves two cases (answer E).
• Answering from the remembered option order - the IBPS key and the CAT/GMAT key are ordered differently, and option wording varies by notification.
• Reporting a non-minimal set in three-statement questions, or stopping at the first sufficient pair without checking whether another pair also works.

Related topics

Inequalities · Statements and assumptions · Statements and conclusions · Critical reasoning: arguments and inferences · Blood relations · Direction sense