The GMAT Data Sufficiency question is the most procedurally distinctive item on the Quant section of the GMAT Focus, and arguably on the whole exam. Every candidate recognises it on sight: a short problem stem, two statements labelled (1) and (2), and the same five-option answer block asking whether the data is sufficient, insufficient, or sufficient only in combination. Yet candidates who can solve the underlying arithmetic perfectly still lose points on these items, because Data Sufficiency is not really an arithmetic test. It is a decision-tree test that uses arithmetic as its substrate. The candidate who treats it as a standard word problem, solving each statement fully before looking at the answer choices, is already spending twice the budgeted time per item. The candidate who treats it as a logic puzzle about the sufficiency of information flips the workflow entirely: parse the stem, isolate the unknown, decide what kind of data would resolve it, and only then read the statements as verdicts rather than as problems to be solved.
This article walks through the stem-first protocol I teach candidates preparing for the GMAT Focus. We will work through the five answer-choice shapes, the two common decision trees, Adam's convention versus a plug-in test for value questions, and the minute-budget triage that lets a candidate leave the section without timing damage. The goal is operational fluency: by the end, a candidate should be able to look at any Data Sufficiency item and know which of the five verdicts is reachable from the stem alone, before either statement has been read.
The five-option answer block, decoded once and for all
Every Data Sufficiency item on the GMAT Focus ends with the same five statements, and the order never varies. The block reads, in essence, that statement (1) alone is sufficient, that statement (2) alone is sufficient, that both together are sufficient, that each alone is insufficient, and that the data is insufficient even when combined. Candidates who memorise the block as five abstract labels waste cognitive load on every item. Candidates who memorise the block as a one-way decision tree treat the answer choices as a search path, not a glossary.
The path is short. Read statement (1). If (1) is sufficient, the answer is choice A, regardless of what (2) says. If (1) is insufficient, the path branches: read statement (2) on its own. If (2) is sufficient, the answer is choice B. If (2) is also insufficient on its own, the path branches a second time: combine the statements. If the combination is sufficient, the answer is choice C. If the combination is insufficient, the path ends at choice E. Choice D — "each alone is sufficient" — only enters the tree at the very end, and only if (1) and (2) each independently give the same answer. Candidates who walk this tree literally mark the answer as they go, and the discipline prevents the most common error: deciding that (1) is sufficient, looking at (2), seeing a trap, and switching the verdict without revisiting the tree.
Notice what the tree does not ask. It does not ask whether (1) is "easier" than (2), or whether (1) "looks like" the more powerful statement. It asks a binary sufficiency question, and only that. In my experience, the candidates who lose points on Data Sufficiency are almost always the ones who allow impressions from one statement to contaminate the verdict on the other. Walking the tree in order, and writing the verdict after each branch, blocks that contamination. Most candidates who adopt the protocol stop making the contamination error within two or three practice sets.
A common sub-error deserves its own mention. When (1) is sufficient, the candidate must commit to choice A without reading (2). The GMAT Focus does not penalise you for reading the second statement, but it does penalise you for the time you spend on it. Train yourself to physically look away from statement (2) once the verdict is locked. The eye will drift back; the discipline is to let it drift back only after the answer has been entered, and only to confirm that the trap answer (the one that says "(1) and (2) together are sufficient but neither alone is") has not tricked you into downgrading your verdict. The five-option block is short. Memorise it, internalise the tree, and the block becomes a procedural checklist rather than a five-way guess.
Stem-first parsing: what the question is actually asking
Before a candidate ever touches statement (1), the stem must be reduced to a single, precise question. Data Sufficiency stems come in three syntactic shapes, and recognising the shape tells the candidate what kind of data will resolve the question.
The first shape is the value question: "What is the value of x?" or "How many litres of solution are in the tank?" Value questions demand a unique numerical answer. Any data that yields two or more possible values for x is, by definition, insufficient. The second shape is the yes/no question: "Is x positive?" "Is quadrilateral ABCD a rectangle?" Yes/no questions demand a definite yes or a definite no. Data that leaves the answer in doubt is insufficient. The third shape is the rarer existence question: "Does there exist an integer n such that..." Existence questions demand a confirmed yes or a confirmed no, evaluated the same way yes/no questions are evaluated.
Reducing the stem to its shape is non-trivial, and candidates who skip this step are the ones who spend two minutes on a statement only to discover the question was a yes/no. The reduction process is mechanical. Read the stem, then read the last line, the one before the statements begin. That last line is almost always the actual question. The arithmetic preamble above it is the setup. Strip the setup, and the question is usually one of: a value of x, a value of an expression in x, an inequality direction, a divisibility claim, or a geometric classification. Each of these maps to one of the three shapes, and the shape maps to a sufficiency criterion.
For value questions, the criterion is uniqueness: would a competent solver arrive at one and only one number? For yes/no questions, the criterion is definiteness: would a competent solver commit to "yes" or to "no" with no hedging? The two criteria are subtly different, and the difference matters. A statement that yields two possible values of x is insufficient for a value question, but it might be sufficient for a yes/no question if both possible values give the same answer to the yes/no. Recognising which criterion is in play is the whole game.
A worked example clarifies. Suppose the stem asks, "Is x positive?" and statement (1) says "x² = 16." A candidate trained on value questions will see two values of x and call statement (1) insufficient. A candidate trained on the shape of the question will note that the possible values of x are 4 and −4; for the question "Is x positive?", 4 yields yes and −4 yields no, so the statement does not produce a definite answer. The verdict is the same in this case, but the reasoning is different, and on harder items the two reasonings diverge. In my experience this usually decides the question: candidates who think in terms of "what would the answer to the stem be?" outperform candidates who think in terms of "what is x?" by a wide margin on yes/no stems.
Adam's convention versus the plug-in test for value questions
Once the stem has been reduced to a value question, the candidate must decide how to evaluate each statement. Two methods dominate the prep literature: the algebraic method and the plug-in test. Algebraic candidates solve the statement for x, and the candidate who can solve the statement is sufficient. Plug-in candidates pick a value that satisfies the statement and see whether the question can be answered; the statement is sufficient if the answer does not change as the candidate picks a second value that also satisfies the statement.
Plug-in is the default for most candidates, and rightly so: it is fast, it is robust against algebraic slips, and it is essentially the only method that works for stems with multiple constraints. Algebra is faster on the rare stem where the candidate can factor the statement in two seconds, and is also faster when the statement is a clean equation in one variable. For the bulk of value questions, plug-in is the workhorse.
The plug-in test has a single, non-negotiable rule: the candidate must plug in two values that satisfy the statement. One value never decides a value question. A statement such as "x is a positive integer" is satisfied by 1, 2, 3, and so on; plugging in 1 tells the candidate what happens when x is 1, but says nothing about what happens when x is 2. If the question is "What is x?", plugging in 1 says "x could be 1"; the candidate must then plug in a second permissible value, say 2, and see whether the question is still answerable. If the answer changes between the two plugs, the statement is insufficient. If the answer is the same for both, the candidate should also try a boundary value — the largest or smallest permissible x, or a value chosen to break the apparent pattern — to confirm the verdict.
Adam's convention is a shortcut within the plug-in method, and it is one of the most useful heuristics on the GMAT Focus. The convention is: when the stem restricts x to positive integers, the candidate should plug in 1 and 2 (and, for yes/no questions, a non-positive integer such as 0 or −1). When the stem restricts x to integers, plug in 0 and 1, plus a negative. When the stem restricts x to real numbers, the convention does not apply directly, and the candidate must use boundary values: the largest or smallest value permitted, and a value far from the apparent centre. The convention is not a law of nature; it is a heuristic that exploits the fact that the test-writer's trap answers are usually triggered by extreme values or by the smallest permissible value. In practice, following the convention catches the trap on roughly four of every five trap-laden items.
Statement (1) and statement (2) as independent verdicts
The single most common error in Data Sufficiency is treating the two statements as a unified problem. The format of the answer block — and the way the test is scored — penalises exactly this. Statement (1) is one mini-problem. Statement (2) is another. The two statements may be combined into a third mini-problem, but only after the candidate has decided that each statement is insufficient on its own. Candidates who read both statements together from the start, in the hope of saving time, almost always overspend time and underperform on accuracy.
The discipline, again, is procedural. Read (1). Decide. If (1) is sufficient, the answer is A and the candidate is done. If (1) is insufficient, set (1) aside mentally — do not carry its information forward into the reading of (2). Read (2) cold. Decide on (2) alone. The carry-forward error, where a candidate remembers a value from (1) and uses it to "evaluate" (2), is a top-three source of lost points on the section. The two statements are independent until the candidate has formally concluded that each is insufficient on its own.
Only after both individual verdicts are insufficient does the candidate combine. The combination step is itself a sufficiency question: would a competent solver, given the union of the information in (1) and (2), commit to a single answer? If yes, the answer is C. If no, the answer is E. The candidate should also verify, at the combination step, that the trap answer D ("each alone is sufficient") is genuinely ruled out. Choice D is the second most common trap on the section, and it usually catches candidates who concluded too quickly that (1) is sufficient. A 30-second sanity check, in which the candidate asks "would (2) alone, with (1) hidden, also yield the answer?" defends against the trap at minimal cost.
A practical note on combining. The combination step is the only step at which the candidate is allowed to use the information from both statements. Even there, the combination is a sufficiency question, not an arithmetic exercise. The candidate should ask: does the combined information force a single answer? If the candidate has to choose between two or more interpretations of the combined information, the answer is E, not C. Choice C is reserved for the case in which the combined information is unambiguous.