What is the GMAT Data Insights section really asking

The GMAT Data Insights section is the third scored component of the GMAT, sitting alongside Quantitative Reasoning and Verbal Reasoning. It measures a candidate's ability to read, interpret, and reason about quantitative information presented in real-world formats: charts, tables, two-part prompts, and short business memos paired with sortable spreadsheets. Twenty questions sit inside a 45-minute window, and the section is scored on the same 60–90 scale used by Quant and Verbal, contributing equally to the candidate's total of 205–805. The framing matters because Data Insights is not a maths test in disguise; it is a literacy test layered over modest arithmetic, and that single distinction reshapes the entire preparation strategy.

Data Insights was introduced when the GMAT moved from the older four-section format to the current three-section, 2-hour 15-minute structure. Its purpose was to measure the data-literacy skills that MBA programmes now expect in classrooms built around case studies, dashboards, and managerial economics. The item families that make up the section, the way they are scored, and the minute-per-question budget that fits inside 45 minutes are the three pillars a candidate has to understand before a single practice set is opened.

The five item families inside the GMAT Data Insights section

The section is built from five recognisable question types, and a strong preparation strategy starts by learning to name each one in under five seconds. The families are: Data Sufficiency, Multi-Source Reasoning, Table Analysis, Graphics Interpretation, Two-Part Analysis. Three of them — Data Sufficiency, Graphics Interpretation, and Two-Part Analysis — also appear in Quant or Verbal on some practice materials, but on the GMAT they are housed inside Data Insights, and the scoring logic and pacing differ slightly from their older placements.

Data Sufficiency remains the most algebra-heavy family. The stem gives a question and two statements, and the candidate judges whether the statements together are enough to answer it. The signature trap is that the answer is almost never the value itself; it is a statement about whether the value can be found. Most candidates who walk into this family cold lose points by recomputing the answer instead of judging sufficiency, and that habit costs roughly 90 seconds per item.

Multi-Source Reasoning presents a small business scenario alongside three tabbed exhibits, only one of which is visible at a time. The test-taker clicks between tabs to gather facts, and three or four questions attach to the same source pack. Because the sources are shared, candidates who skim once and answer slowly usually outperform candidates who read carefully and rush the second item in the pack.

Table Analysis attaches a sortable spreadsheet to three open-ended questions. Sorting by a column is the single most useful first move; most questions stop being ambiguous once the table is reordered around the right variable. Graphics Interpretation shows one chart, usually a scatter, stacked bar, or line graph, and asks two to three questions about ratios, percentages, or trends. Two-Part Analysis is the trickiest family: a single prompt feeds two related quantities that have to be selected from a five-by-six grid of options, one from each row, and the pair has to satisfy the stem.

The order in which these families appear is not fixed, and the adaptive algorithm does not signal which one is coming. Candidates who have pre-decided a default reading protocol — stem first, exhibit second, options third — adapt faster to whichever family the test serves up next.

How the section is scored on the GMAT

Data Insights contributes its own 60–90 score to the total, and that score is the third leg of the 205–805 total. The five item families are not weighted identically. Data Sufficiency typically accounts for the largest share of questions on most test forms, while Two-Part Analysis is the smallest family by count but the highest by time cost per item. A practical way to read the weighting is to think in minutes, not question counts: a section that gives you 6 Two-Part items and 4 Data Sufficiency items still rewards Data Sufficiency in raw score terms, because each correct DS answer contributes one raw point and each Two-Part contributes one raw point, but a single missed Two-Part question erases roughly 2.5 minutes of opportunity cost.

The section is adaptive at the item level: each question is chosen based on your answers to the previous ones. Each section is adapted on its own, so your Data Insights performance does not set the difficulty of the Quant or Verbal items.

Raw-to-scaled translation, in practical terms

Two facts make the raw-to-scaled jump less mysterious. First, the section has only 20 questions, so every miss is visible in the 60–90 score. Second, the score depends on both the number of correct answers and the difficulty of the questions answered, and unanswered questions are penalised. That structure is why pacing evenly through the section matters more than sprinting the first ten items; you want to be fresh enough to answer the late items at full accuracy.

For most candidates, a target of 16 to 18 raw correct translates to a scaled band of 81 to 87. A candidate who misses only Two-Part items usually lands lower than a candidate who misses the same number of Data Sufficiency items, because the engine treats each family as a partial-credit signal. The practical takeaway: protect the late-section Two-Part items even at the cost of an earlier Graphics Interpretation miss.

The minute-per-question budget across the 45-minute window

Twenty items in 45 minutes gives an arithmetic budget of 2 minutes 15 seconds per question, but no serious preparation strategy treats that average as the working pace. The honest breakdown is closer to 1 minute 30 seconds for the shortest family, 1 minute 45 seconds for Data Sufficiency, 2 minutes 30 seconds for Table Analysis, and 3 minutes for a Two-Part Analysis item. Multi-Source Reasoning sits somewhere between 2 and 2.5 minutes per question, depending on how the three tabs are linked.

Building that budget into a preparation plan means practising each family under its own timer, not under the section-wide average. A candidate who drills 20 mixed questions in 45 minutes repeatedly is practising the wrong skill: pacing across families. A candidate who drills four Data Sufficiency items in seven minutes is practising the right skill: holding the family-specific pace under exam conditions.

Pacing traps that cost more than one question

Three pacing traps recur again and again. The first is the "Data Sufficiency deep dive", where the candidate starts computing the answer once they spot the obvious algebraic move. The second is the "Two-Part grid scan", where the candidate reads the five rows of options before re-reading the stem. The third is the "Multi-Source tab click", where the candidate flips tabs in alphabetical order instead of following the chain of references the prompt implies. Each trap costs between 30 and 60 seconds on its first appearance and 90+ seconds once it compounds across an item cluster.

How the question types differ in cognitive load

Cognitive load is the right frame for distinguishing the families, because the GMAT does not publish a difficulty rating for each item. A Data Sufficiency item rarely asks for hard arithmetic; it asks for a clean judgement about whether the two statements, alone or together, license a conclusion. The cognitive load is logical, not computational. A Graphics Interpretation item asks for an estimate from a visual; the load is perceptual. A Two-Part Analysis item combines a logical structure with a constrained search across a 5-by-6 option grid; the load is procedural, and it punishes candidates who have not internalised a row-by-row elimination method.

Table Analysis items are the most under-rehearsed family in most preparation plans. The visible artefact is a spreadsheet with sortable columns, and the work is to identify which column or sort order turns the question into a one-line lookup. The cognitive load is investigative, and candidates who have not seen a real sortable grid in practice will spend the first 30 seconds figuring out how the click-and-sort interaction works.

Reading the stem before the exhibit: a universal first move

One habit transfers cleanly across all five families: read the stem to its end before looking at the exhibit. The stem contains the question, the requested units, and the binding constraints. Candidates who read the exhibit first tend to anchor on the most visually striking feature of the chart or table and then search for a question that matches it, which is the reverse of what the section rewards. Reading the stem first lets the candidate enter the exhibit with a specific target, and that targeted reading is faster and more accurate.

A preparation strategy that matches the section's structure

Effective preparation for Data Insights is built in three layers, and skipping a layer is the most common reason candidates plateau around a scaled score of 78. The first layer is family recognition: a candidate should be able to name each of the five families within five seconds of seeing the stem. The second layer is family-specific technique: each family has a one- or two-step protocol that, when followed, turns the item into a mechanical exercise. The third layer is section-level pacing: integrating the five family-specific techniques into a single 45-minute run without leaking time at family boundaries.

A six-week preparation plan that targets a scaled score of 84 or higher typically allocates the first two weeks to layer one, weeks three and four to layer two, and the final two weeks to layer three with a weekly full-section simulation. Candidates who collapse all three layers into the first three weeks tend to be confident on day 20 and demoralised on day 35, because they have not built the late-section stamina that the adaptive engine rewards.

Drills that move the score more than review does

Two drills outperform passive review for this section. The first is a five-by-five grid drill, where the candidate solves 5 Data Sufficiency items in 8 minutes, then 5 Graphics Interpretation items in 8 minutes, then 5 Table Analysis items in 12 minutes, then 3 Two-Part items in 9 minutes, and finally 2 Multi-Source items in 8 minutes. The second is an error-tag drill, where the candidate logs every miss across a week of practice, tags each miss as either a recognition failure, a technique failure, or a pacing failure, and then targets the dominant tag in the next week.

Common pitfalls and how to avoid them on test day

Five pitfalls account for most of the point loss on Data Insights, and each has a concrete counter-move. Candidates who memorise the counter-moves in advance walk into test day with a default response to the most common ways the section can derail them.

• Computing the answer on Data Sufficiency. The question asks whether the statements are enough, not what the value is. The counter-move is to train the habit of writing the verdict on a scratch line before reaching for the calculator.
• Sorting the wrong column on Table Analysis. The first instinct is to sort by the column named in the question, but the right column is often one that the question implies, not one it names. The counter-move is to identify the variable that, once ordered, makes the answer a single cell, and then sort by that variable.
• Reading the chart instead of the question on Graphics Interpretation. Candidates lose points by inferring trends the chart does not support. The counter-move is to mark the two endpoints the stem names, then read the question a second time before selecting.
• Ignoring the second answer on Two-Part Analysis. The two answers are linked, and selecting the correct first row without confirming the second row is a reliable way to lose both points. The counter-move is to treat the two answers as a single object and to verify the pair before submitting.
• Tab-hopping on Multi-Source Reasoning. Flipping between the three tabs without a plan burns 20 to 30 seconds per question. The counter-move is to scan the prompt for the order in which the tabs are referenced, then follow that order.

Comparing the five item families at a glance

The table below summarises the family-level trade-offs a candidate has to manage. Times are working times, not arithmetic averages, and they assume a candidate who is targeting a scaled band of 81 to 87.

Scroll the table sideways →

How Data Insights interacts with Quant and Verbal on the GMAT

On the GMAT, you choose the order of the three scored sections yourself before the test starts. The order matters less than the fact that all three sections share the same 60–90 scale. Schools reading the score report see a candidate's profile as three numbers that have to make sense together. A candidate with a Quant 85, Verbal 81, and Data Insights 74 is sending a signal that data literacy is the weakest of the three skills, and admissions committees read that as a hint that case-method coursework will land harder for the candidate than for a peer with a balanced 80–80–80 profile.

The interaction also matters inside the test. The adaptive engine uses your answers within a section to choose the difficulty of the next item in that section. Candidates who treat Data Insights as a warm-down section tend to underperform on it, and because Data Insights carries the same weight as Quant and Verbal in the total, a serious preparation strategy treats all three sections as equal-weight events.

Building a personal preparation plan around the section

Personalising the plan means starting with a diagnostic that produces a family-by-family accuracy profile, not a single section score. A candidate who scores 75 overall but is 90 percent accurate on Graphics Interpretation and 50 percent accurate on Two-Part Analysis should spend the first week on Two-Part, not on a general review. A candidate who scores 75 overall but is uniformly 75 percent accurate across all five families should spend the first week on pacing, because uniform accuracy usually masks a leaking time budget.

The diagnostic also reveals the order in which the candidate should attack the families in the final two weeks of preparation. Candidates cannot choose the order in which item families appear, but they can decide in advance which families get a quick, triaged guess when time runs short and which get the full time budget. That plan is a tactical advantage that pure content review does not provide.

Tracking progress without obsessing over the score

The most useful progress metric for this section is the rolling two-week accuracy rate per family, expressed as a percentage, paired with the average working time per family in seconds. A candidate whose accuracy is rising while the working time is stable is improving. A candidate whose accuracy is rising but whose working time is also rising is over-learning items and will leak time on test day. A candidate whose accuracy is flat but whose working time is falling is internalising protocols but not yet hitting the harder items in each family. Watching the two numbers move together is a more honest signal than watching the scaled score move week to week.

What the item families reward, and how to read your progress

Data Sufficiency in Data Insights is not limited to pure algebra. Many stems are business or data questions ("Did revenue exceed budget in the third quarter?"), and the first move is still to decide whether the question asks for a value, a yes/no answer or a comparison. The same habit helps across the section: name the item family, name the shape of the question, then act.

• Multi-Source Reasoning: when a question asks which statements are supported, mark each one supported, contradicted or undetermined before you look at the answer choices.
• Graphics Interpretation: statements often pair a true observation with a claim the chart does not support. Check quantifiers such as most, all and some, and do not accept a cause the chart cannot show.
• Two-Part Analysis: when the two parts are linked by a trade-off (cost against speed, risk against return), write the trade-off down first; picking each column in isolation is the usual source of an inconsistent pair.
• Table Analysis: sort by the column a statement names before you judge it.

Reading progress without overreacting

With only 20 questions, Data Insights scores move noticeably from one practice set to the next. Judge progress on the trend across several sets rather than on the latest one. Time is often the earlier warning sign: a family on which you keep running long, even when you get it right, is usually the next family where accuracy will slip. A single fully reviewed practice set, with notes on which items took too long, which you got right for the wrong reason and which you misread, teaches more than several sets taken without review.

Conclusion and next steps

The GMAT Data Insights section rewards a specific kind of literacy: the ability to read a question, enter a chart or table with a target in mind, and judge whether the evidence licences a conclusion. The five item families, the 45-minute window, and the equal-weight 60–90 scale together form a section that is genuinely different from Quant and Verbal, and the candidates who treat it as a separate preparation project — with its own diagnostic, its own drills, and its own pacing targets — are the candidates who post balanced score reports. The single most useful next move for a candidate at the start of preparation is a family-by-family diagnostic that produces a working-time-and-accuracy profile, because that profile is what every later decision will be built on. TestPrep Europe's diagnostic assessment is a natural starting point for candidates building a sharper preparation plan around the GMAT Data Insights section.

Frequently asked questions