- Data sufficiency now lives in Data Insights, so a miss costs a third-of-the-total section point, not a Quant point.
- The question is never ‘what is x’; it is ‘is x pinned down,’ and solving for x is the most expensive habit a strong quant candidate brings in.
- Decide what sufficient would require before you read either statement, then evaluate each statement alone before combining.
- A definite no is sufficient; a ‘maybe’ is not; and a variable is not a positive integer unless the stem says so.
- The five answer choices never change, so the procedure can be fixed and the time per question can drop to 1:30.
You are strong at Quant, you have been for years, and you keep missing data sufficiency questions that you could solve in your sleep if someone just asked you to solve them. That is the problem in one sentence: data sufficiency does not ask you to solve, and the solving habit that makes you good at everything else is exactly what costs you here.
The format moved. On the Focus Edition, GMAC (mba.com) places data sufficiency in the Data Insights section rather than Quant, alongside multi-source reasoning, table analysis, graphics interpretation and two-part analysis, in a 20-question, 45-minute section that carries one third of your 205 to 805 total (GMAC exam structure). The question format and the five fixed answer choices did not change. Neither did the logic, and the logic is five rules.
What a data sufficiency question is actually asking
A data sufficiency item gives you a question, two numbered statements, and five answer choices that are the same on every item: statement one alone is sufficient but two is not; two alone is sufficient but one is not; both together are sufficient but neither alone; each alone is sufficient; both together are still not sufficient. Your job is to pick the letter. You never report a value.
The questions come in two kinds. A value question asks for a specific number: what is x, or how many units did the plant produce in March. Sufficient means exactly one value. A yes/no question asks whether something is true: is x greater than 5, or did revenue increase. Sufficient means the answer is definitely yes or definitely no. A definite no is sufficient. “Sometimes yes” is not.
Hold that distinction. Almost every rule below depends on it.
Rule one: decide what sufficient means before you read the statements
Before your eyes touch statement one, rewrite the question in your own words and write down what would have to be true for a statement to be sufficient. For a value question, that often means rearranging: if the question asks for the value of 3x + 6, sufficient means knowing x + 2, which is weaker than knowing x. For a yes/no question, it often means simplifying the condition: “is x² > x” is really “is x > 1 or x < 0,” and a statement that puts x in either region is sufficient.
Strong quant candidates skip this step because they can “see” the problem. Then they evaluate the statements against the original question rather than the simplified one, and they end up proving more than they needed to, which costs time, or missing that a weaker piece of information was enough.
Rule two: each statement alone, with the other physically hidden
Evaluate statement one by itself, write your verdict, then evaluate statement two by itself as if statement one did not exist. The hiding is literal. Cover the other statement with your hand on the noteboard, or scroll it out of view, because the single most common data sufficiency error is carrying a fact from statement one into your reading of statement two without noticing.
The procedure produces a two-letter record on your noteboard: S or N for each statement. Only if both are N do you combine. If you have S and N, your answer is already determined, and any further work is waste.
Write the letters down every time.
Rule three: a definite no is sufficient
Yes/no questions produce the most reversed verdicts. A statement that proves the answer is no has answered the question, which makes it sufficient. Candidates who are solving rather than deciding treat “no” as a failure and mark the statement insufficient.
The fix is to write the two words “yes” and “no” under the question and strike one as each statement settles the matter. If a statement leaves both standing, it is insufficient. If it strikes one, it is sufficient, whichever one it struck.
Rule four: do not assume what the stem did not say
Unless the stem says so, a variable can be zero, negative, a fraction or an irrational number. “Integer” does not mean positive. “Positive” does not mean integer. “Number” means anything. Strong quant candidates carry assumptions from the problem-solving world, where the answer choices quietly constrain the variable, into data sufficiency, where nothing constrains it but the stem.
The practical check is a short list of test cases you run against any statement that looks sufficient too quickly: zero, a negative, a fraction between 0 and 1, and a large number. If any of them produces a different answer, the statement is not sufficient.
Test zero first. It breaks the most arguments.
Rule five: solve only enough to decide, then stop
The time you save on data sufficiency pays for the rest of the section, and the saving comes from stopping early. Once you know a statement produces exactly one value, you do not compute the value. Once you find two cases with different answers, you do not look for a third. Once you have S and N, you do not combine.
This is the rule that strong quant candidates resist, because finishing the computation feels like checking your work. It is not. The check is whether the logic was right, and the computation does not test the logic.
The trap list
The five rules cover the logic. The traps below are the specific ways the test gets a strong candidate to break one of them, and they repeat often enough that you should recognize each by name.
The leaked statement. Information from statement one shapes your reading of statement two. Fix: hide the other statement, write S or N before uncovering.
The hidden quadratic. A statement like x² equals 16 or |x| equals 4 gives two values, which is insufficient for a value question unless the stem restricts the sign. Fix: when you see a square or an absolute value, write both roots before deciding.
The positive-integer assumption. The stem says “number” or “integer” and you treat it as a positive integer. Fix: test zero, a negative and a fraction before marking sufficient.
The combined-too-early habit. You read both statements together because the problem looks hard, and you can no longer tell which one did the work. Fix: the two-letter record, always.
The disguised equivalence. The two statements say the same thing in different clothing (one says 2x plus 4y equals 10, the other says x plus 2y equals 5), so if one is insufficient the other is too, and combining adds nothing. The answer is never “both together” when the statements are equivalent. Fix: if the statements look related, simplify both before evaluating.
The “C trap.” Both statements together obviously work, and the question is designed so that you never check whether one alone did. The test rewards candidates who check each alone first. Fix: rule two, without exceptions.
The no-is-insufficient reflex. A statement proves the answer is no and you mark it insufficient. Fix: the yes/no strike-through.
The arithmetic you didn’t need. You compute the value, make an arithmetic slip, decide the statement “doesn’t work,” and mark it insufficient when the logic was fine. Fix: rule five; the value is never the question.
That is eight traps. Most candidates have two or three favorites, and the error log will tell you which. Once you know yours, the fix is a single written habit on the noteboard, and the misses stop within a couple of weeks.
The procedure on one line
Here is the whole method as you would write it at the top of your noteboard on test day.
Question: value or yes/no. Sufficient means: (write it). Statement one alone: S or N. Statement two alone, with one covered: S or N. Combine only if N and N. Test zero, negative, fraction before any S. Stop when decided.
That is 30 words, and it is the difference between a 1:30 question and a 2:45 question. Spend a week running it on every data sufficiency item you touch, untimed, until the letters appear on the noteboard without your deciding to write them. Then time it. Most candidates who do this see the question type go from their least reliable to their most, because it is the only type on the test with a fixed procedure and fixed answer choices, and fixed procedures are what a strong quant mind is good at.
Your quant ability was never the problem. The solving reflex was.
What to do this week
- Write the one-line procedure on an index card and keep it beside you for every data sufficiency set this week.
- Do 15 data sufficiency questions untimed tonight, covering the unused statement with your hand, writing S or N for each before uncovering.
- For every question you mark sufficient, run zero, a negative and a fraction against the statement before you commit.
- Log each miss against the eight traps by name and circle the one that appears most.
- On Saturday, time a set of 15 at 1:45 each and compare your accuracy to the untimed set; if it dropped, you are solving again.
Questions people ask
Why did GMAC move data sufficiency out of Quant?
The Focus Edition rebuilt the sections so that Quant is problem solving only and Data Insights holds the question types that test reasoning about information. Data sufficiency fits the second category: it asks whether information is enough, not what the answer is. The question format and the five answer choices are the same as before.
How long should a data sufficiency question take?
Target 1 minute 30 seconds to 2 minutes, against a section average of 2:15. The surplus is deliberate; it pays for multi-source reasoning prompts elsewhere in the section. If you are regularly over 2:30, you are solving when you should be deciding.
Should I plug in numbers or work algebraically?
Both, in a fixed order. Try to settle the statement algebraically in 20 seconds; if you cannot, plug two deliberately different cases (a positive and a negative, an integer and a fraction, zero and not zero) and look for two different answers. Two different answers means not sufficient, and you are done. Plugging is a tool for proving insufficiency fast, not for finding the value.
Do I need to combine the statements every time?
No. You combine only when each statement alone is insufficient. If statement one is sufficient and statement two is sufficient, the answer is D and combining is wasted time. The most common procedural miss is combining too early, which also lets information from one statement leak into your reading of the other.
Are data sufficiency questions harder in Data Insights than they were in Quant?
The logic is identical. What changed is the neighborhood: they now sit between table analysis and multi-source reasoning questions, so you switch modes more often inside a section. Candidates who miss them in Data Insights usually did not have a fixed procedure to begin with, and the mode-switching exposes that.