Data Insights is a third of your score. Here is the method.

Your Quant is in the 80s, your Data Insights is sitting at 72, and the gap is costing you more total points than anything else on your score report. The fix is a method per question type, not more math, because Data Insights is not testing whether you can compute. It tests whether you can decide what matters and stop.

The section is one third of your 205 to 805 total, equal in weight to Quant and Verbal. GMAC (mba.com) sets it at 20 questions in 45 minutes, with an on-screen calculator available only in this section, across five question types: data sufficiency, multi-source reasoning, table analysis, graphics interpretation and two-part analysis (GMAC exam structure). That is 2 minutes 15 seconds a question on average, and the average is misleading, which is the first thing to fix.

Why quant-strong candidates stall at 72

Candidates from finance and consulting arrive with a habit that serves them well in Quant and badly here: when they see numbers, they solve. In Data Insights, solving is frequently the wrong first move. The question usually asks whether something is true, whether a value can be determined, or which of several statements the data supports. Computing a precise figure when a comparison would do costs 40 seconds a question, and over 20 questions that is the whole section’s slack.

The second habit is reading everything. A multi-source reasoning prompt has two or three tabs of text, tables and emails. Reading all of it before looking at the questions feels thorough and wastes two minutes per prompt. The questions tell you which tab matters.

The third habit is the one nobody admits to: treating the section as an afterthought in study. Quant gets the hours because Quant is comfortable. We flip that ratio for most clients, and the section typically moves faster than either of the others because the gap is method, and method is learnable in weeks.

The one rule that covers all five types

Read the question before the data. That is the whole method in one sentence, and every type-specific move below is a version of it.

The data in this section is deliberately dense: tables with eight columns, or three tabs of correspondence about one policy. Almost every question uses a small fraction of what is shown. If you read the stem first, you know which fraction. If you read the data first, you carry all of it in working memory while you look for the question, and working memory under a clock is unreliable.

Practice this on every single question until it is automatic. Stems come first.

Data sufficiency: decide, do not solve

Data sufficiency asks whether the information in two statements is enough to answer a question, alone or together. The five answer choices are always the same, in the same order, which means the format rewards a fixed procedure.

Before you look at either statement, rewrite the question in your own words and decide what “sufficient” would require. If the question asks for the value of x, sufficient means exactly one value. If it asks whether x is greater than 5, sufficient means a definite yes or a definite no; a definite no is sufficient. Then evaluate statement one alone, write a letter, evaluate statement two alone, write a letter, and only if both fail alone do you combine them.

The traps are consistent. Carrying information from statement one into your reading of statement two is the most common, and the fix is physical: cover one statement with your hand or scroll it off screen. Solving for the value when the question is a yes/no is the second, and it costs time rather than accuracy. The third is assuming a variable is a positive integer because the problem feels like it should be; unless the stem says so, zero, negatives and fractions are in play.

Budget 1 minute 30 seconds to 2 minutes here. The surplus pays for multi-source reasoning. We have a longer guide on the logic rules and the trap list at /gmat/guides/gmat-data-sufficiency-logic/.

Table analysis: sort once, then answer

Table analysis gives you a sortable table and a set of statements to mark true or false (or yes or no, or would help or would not). The move is to read all the statements first, decide which column each one depends on, and sort by that column once. A table sorted by the right column turns a hunt into a glance.

Watch the units and the headers. A column labeled “change from prior year, %” is not the same as a column of values, and the statement will be written to catch a reader who treats them alike. Watch for statements that need a ratio of two columns; those are the ones to leave for last, with the calculator.

Each set is scored as one unit. Finish it.

Graphics interpretation: name the axes before the question

Graphics interpretation shows a chart (scatter, bar, line, bubble, occasionally a flowchart or Venn diagram) with one or two sentences that have dropdown blanks. Before you read the sentence, say the chart’s two axes and its unit out loud or on the noteboard. “Horizontal is year, vertical is revenue in millions, two series.” That 10-second habit eliminates the most common miss, which is reading the right point off the wrong axis.

Then read the sentence with the blanks, decide which dropdown you can fill from a single reading and which needs a comparison, and do the single readings first. Dropdown options are usually spaced widely enough that an estimate is fine; if two options are within a few percent of each other, that is your signal to use the calculator rather than the eye.

Target 1:45 to 2:00 per question. Bank these at pace.

Two-part analysis: find the link between the two columns

Two-part analysis presents a short scenario and a table with two columns of answer choices that share one list of options. You pick one option per column. It can be quantitative (two values that satisfy two conditions) or verbal (a strengthening and a weakening statement for an argument).

The key is that the two parts are almost always linked. In the quantitative version, solve for the relationship between the two unknowns first, then test option pairs against it; do not solve each column independently. In the verbal version, prephrase what each column needs before reading options, the same way you would in critical reasoning. Candidates who read options first and try to fit them into columns take twice as long and get caught by options that fit one column and sound right for the other.

Budget 2:30. It is the type most likely to run long, and it is a fair place to flag and return.

The time budget across the section

Twenty questions in 45 minutes averages 2:15, but you should plan on an uneven spend. The table below is the budget we hand clients in week 2.

Type counts vary by form, and GMAC does not publish a fixed distribution, so treat the counts as what we typically see rather than a rule. The pattern that matters is this: data sufficiency and graphics interpretation are where you run ahead of the clock, and multi-source reasoning and two-part analysis are where you spend the surplus.

Check the clock after question 7 and question 14. Two checks are enough. If you are more than two minutes behind at 14, the next multi-part set you are unsure of gets a flag and a best guess, not a fight.

Multi-source reasoning: a worked example

Multi-source reasoning is the type that decides the section for most candidates. A prompt shows two or three tabs (text, a table, an email thread, a policy memo) and then asks two or three questions about them, each of which may itself have three yes/no parts. The reading is heavy and the questions are cheap once the reading is done, so the time budget has to be set across the whole prompt, not per question.

Here is an original example in the shape the test uses.

Work it the way the method says. Read the three question stems first. Question 1 depends on the fourth-quarter column and the 80 threshold. Question 2 depends on the same column and the 90 threshold. Question 3 depends on the email. Now you know the memo gives two rules, the table gives one column that matters, and the email gives one exception. You never need the third-quarter column at all.

Question 1. Alder’s fourth-quarter score is 79, below 80, so the memo requires a plan: yes. Birch is at 93: no. Cedar is at 82, above 80, so the memo does not require a plan, and the email’s provisional-score issue does not change that; a provisional 82 is still not below 80 on the information given: no. The trap is Cedar, where the email makes you want to say yes because something feels unresolved. The question asks what the tabs establish. They establish 82.

Question 2. Exempt means 90 or above in the fourth quarter. Birch at 93 and Dogwood at 90 qualify; the “or above” wording includes 90. Elm’s fourth-quarter score is 86, and its third-quarter 90 is the trap for anyone who read the table before the stem. Answer: Birch and Dogwood.

Question 3. Cedar’s score is 82. For its audit status to change, the recalculated score would need to reach 90 from a provisional 82, and nothing in the tabs rules that out. The statement says “could change,” not “will change.” True. Candidates who answer false are usually arguing with the likelihood, and the question did not ask about likelihood.

Total reading, if you read stems first: the memo’s two rules, one column of five numbers, and one email exchange. That is about 90 seconds of reading for three questions. Candidates who read all three tabs top to bottom before the first stem spend three minutes reading and then re-read the table for each question anyway.

How to practice this in four weeks

Week one is data sufficiency alone, 20 questions a night at 1:30 to 2:00 each, with the five-letter procedure written on the noteboard until you stop needing it. Week two is table analysis and graphics interpretation mixed, aiming for 1:45 and checking the axis habit every question. Week three is two-part analysis and multi-source reasoning, untimed for the first two nights so the method sets before the clock does, then timed. Week four is full 20-question sections at 45 minutes, twice a week, with the stems-first rule enforced on every question.

Log every miss by type and by cause. The causes in this section are nearly always one of four: read the data before the stem, solved when a comparison would do, misread a header or axis, or mistyped a calculator entry. If the same cause shows up five times in a week, that is your pattern, and that is what the next week’s sets are for.

Most candidates who come in at 72 and do this for four weeks see the section move more than any other. We will not promise a number. We will say that the gap you have is the kind that closes.

Rick’s note: I ask every Data Insights client to say the axes out loud for the first two weeks. It feels silly. The candidates who keep doing it quietly at the test center are the ones whose graphics-interpretation misses go to zero, and I have stopped apologizing for the silliness.

What to do this week

1. Pull your Enhanced Score Report, or your last practice-test breakdown, and write down the Data Insights score next to Quant. The gap is the budget for this work.
2. Do 20 data sufficiency questions tonight with the two statements physically covered one at a time. Log each miss by cause.
3. On every Data Insights question this week, read the stem before the data and say the chart’s axes before reading a graphics-interpretation sentence.
4. Time one 20-question Data Insights set this weekend at 45 minutes and check the clock at questions 7 and 14.
5. Sort the misses into the four causes (data before stem, solved instead of compared, misread header or axis, calculator typo) and pick the one that repeats.