5 GRE Data Analysis item families and the 90-second triage for each

GRE Data Analysis is the strand of the Quantitative Reasoning section where the test stops rewarding arithmetic and starts rewarding graph literacy. About 10 to 12 of the 27 unscored and scored Quant items on a single sitting live inside this strand, and they are almost always the items candidates score highest on in practice yet underperform on test day. The reason is rarely arithmetic weakness. It is the way a chart, a two-way table, or a five-line stem forces the test-taker to redefine the problem in real time, under a clock, with a pencil that feels too slow.

This article is built around that exact problem. It walks through the three item families the test uses, the reading moves that turn a dense table into a 45-second question, and the habits that separate a 158 from a 165+ on the Quant section. Every example, every pitfall, and every heuristic below is anchored to the data-analysis strand of GRE Quant and to the preparation style that tends to lift scores fastest.

The three item families inside GRE Data Analysis

Candidates preparing for the GRE General Test often hear "Data Analysis" used as a single label, but the strand actually contains three distinct item families, and the way you read the stimulus changes between them. A student who learns one rhythm for the whole strand loses minutes on items they should be polishing in under 90 seconds.

The first family is the data-interpretation set: a shared stimulus, usually a bar chart, line graph, or two-way table, followed by three discrete questions. The second family is the single-question graphical item: a small chart or table, one question, no follow-ups. The third is the quantitative comparison in graphical clothing: two columns, A and B, supported by a visual that gives away far more than the numbers alone. Each of these three behaves differently under timed conditions, and each demands a different first glance.

Data-interpretation sets are the highest-leverage items on the section. Because three questions share one stimulus, the cost of misunderstanding the chart is three wrong answers, not one. Test makers know this and design the charts so that the first question is the easiest and the third is the most synthetic. If a candidate can read the y-axis label, identify the right series, and avoid a rounding trap, the first question is almost free. Most of the points lost in this family are lost on questions two and three, where the test asks the student to combine two pieces of the chart or to project beyond it.

Single-question graphical items are the filler, the air traffic of the section. They are quick, contained, and rarely tricky in a conceptual sense. A candidate who knows how to read a pie chart, a boxplot, or a five-row frequency table can clear these in 60 to 75 seconds. The risk here is not difficulty but attention: the test maker plants one small misread — a misread axis, a different denominator, an off-by-one bar — and waits for the tired eye to grab it.

Quantitative comparisons dressed as data items are the third family and the one that confuses most candidates in the diagnostic stage. The visual looks like a regular bar chart, but the answer choices are the familiar A/B/C/D quartet. The stem of a graphical QC is almost always doing more work than the chart. If the chart is a five-bar graph and the question asks about a ratio of two specific bars, the candidate who reads the numbers off cleanly usually wins the item. If the candidate starts trying to estimate percentages from eyeballed angles, the item is already lost.

Reading the stimulus in 20 seconds: a four-step skim

Every data-analysis item begins with a stimulus that takes 15 to 25 seconds to interpret. A senior tutor's job is to compress that window to a reliable four-step skim that any candidate can repeat under pressure. Skipping any one of these four steps is the single most common cause of a 'stupid mistake' on a graph item.

Step 1: read the title and the axis labels before the numbers. The title tells the candidate what is being measured. The y-axis label tells the unit. The x-axis label tells the categorical dimension. A candidate who starts hunting for the largest bar without reading the unit is the candidate who picks the answer for a different question. On a bar chart titled 'Average monthly rainfall in millimetres, 2010–2020,' the number '120' means 120 millimetres, not 120 inches and not 12 centimetres. Reading the unit first eliminates a whole category of careless errors before the pencil touches the page.

Step 2: locate the legend, the series labels, and the source line. A line graph with two series labelled 'urban' and 'rural' is two questions, not one. A pie chart with five slices labelled only by colour is a test of whether the candidate can read the slice labels at all. A footnote line that says 'values rounded to the nearest 0.1' is the test maker's hint that one of the answer choices will collapse if the candidate tries to compute to a higher precision than the chart supports.

Step 3: identify the most extreme data point in each series. For a bar chart, that is the tallest and shortest bar. For a line graph, the highest and lowest point on each line. For a two-way table, the maximum and minimum cell. The extreme point is the anchor: most of the questions in the set will refer back to it, either directly or by comparison. A candidate who has already located the maximum bar in 20 seconds answers the easy first question immediately and uses questions two and three as a confidence check.

Step 4: mark any visual trap. A broken y-axis, a logarithmic scale, a pie chart where the slices do not sum to 100, a bar chart with a missing category, a table with a row marked 'n/a' — these are the four or five physical features that change the meaning of the data. Candidates who have learned to scan for them on the first glance spend 30 to 45 seconds less per set and lose far fewer items to what the test makers call 'graph illiteracy'.

Item-by-item triage: how the test makers sequence a data set

A data-interpretation set is not three questions of equal difficulty. It is a 1-2-3 ramp: question one tests reading, question two tests reading plus a single arithmetic step, question three tests reading plus arithmetic plus a synthesis that the chart does not directly show. Recognising that ramp is the single most useful piece of sequencing knowledge a candidate can carry into the section.

Question one in a typical set is a direct-lookup. 'In 2018, approximately what was the ratio of urban to rural households with internet access?' The chart shows the bars, the candidate reads the heights, computes the ratio, and moves on. In my experience, candidates who skip the first question and start at question two do not save time — they lose the warm-up effect and then misread question two's chart in exactly the same way they would have misread question one's. The warm-up question exists for a reason, and a senior tutor's advice is to use it.

Question two usually adds a layer. Either the candidate must combine two bars from the same series, or they must read a value that sits between two gridlines and apply linear interpolation, or they must translate a percentage into a count. The arithmetic is light — a 15 to 30 second calculation — but the read must be precise. This is the question where the rounding trap lives. A pie chart where the slice is labelled '14%' should never be treated as 14.000 per cent. The test maker will plant a '14.4%' answer choice to punish the candidate who rounded too aggressively.

Question three is the synthesis. The test asks something the chart does not directly show: 'If the trend from 2015 to 2019 continues, in what year will the urban line first exceed twice the rural line?' or 'What is the difference between the average of series X and the median of series Y?' The candidate has to do at least one operation the chart does not give, usually a projection or a cross-series comparison. This is the question where 60 to 90 seconds of pencil time is the right budget, and where the disciplined candidate stops trying to compute to four significant figures and accepts the answer within the chart's visible precision.

Common pitfalls and how to avoid them

• Skipping the first question in a set to 'save time for the harder ones.' In practice this costs 30 to 60 seconds of additional reading later and one missed warm-up item.
• Rounding intermediate values to whole numbers when the chart supports one decimal. The answer choices are designed to punish this; pick the choice that matches the chart's own precision.
• Ignoring the 'approximately' in the stem. 'Approximately' is permission to estimate, not permission to round to the nearest integer. Keep one decimal in the estimate.
• Computing on a broken y-axis as if the axis were linear. A bar chart where the y-axis starts at 47 instead of 0 makes a 1-unit bar look like a 3-unit bar. Always scan the y-axis baseline first.
• Forgetting the legend on a multi-series line graph. Candidates regularly answer the question for series A when the stem asked about series B. A 20-second legend check kills this error class.

Single-question graphical items: the speed math of the section

Single-question graphical items are the air traffic of the section: short stimuli, single questions, low conceptual ceiling, and a tight time budget. The diagnostic question for a candidate is not 'can I get these right in isolation' but 'can I get them right under a 75-second budget, three at a time, after a 12-minute verbal block'. That is the real test.

The item types in this family include pie charts, boxplots, scatter plots with a single trend line, frequency tables with row and column totals, and small bar charts with one series. A candidate preparing for the GRE General Test should be able to clear any of these in 60 to 75 seconds once the four-step skim is internalised. The arithmetic is almost never the bottleneck; the read is.

On pie charts, the most common error is converting a slice to its count using the wrong whole. A pie chart showing 'expenditure by category' is not the same as 'revenue by category' and the total at the bottom of the chart is the only safe whole. Candidates who reach for the familiar 360-degree total without checking the legend lose a point every two to three sittings. A safe rule: write the total beside the chart before you compute.

On boxplots, the test tends to test the five-number summary directly: minimum, first quartile, median, third quartile, maximum. The visual is dense, and candidates often mistake the whisker for the quartile. A candidate who has memorised the boxplot anatomy — box is Q1 to Q3, line inside the box is the median, whiskers extend to the min and max — clears these in under 60 seconds. A candidate who has to reason it out from scratch on test day spends 90 to 120 seconds and burns the next item's clock.

On scatter plots, the test is almost always about the trend line, not the data points. The candidate is asked to predict, compare, or interpret a slope. The two skills that pay off here are (1) reading two points on the line and using them to compute the slope, and (2) recognising that a scatter plot with no fitted line is asking the candidate to estimate the line, not to find it precisely. Both are 30-second operations once the candidate is fluent in slope as rise over run.

Quantitative comparisons dressed as data items

Quantitative comparison questions that arrive with a graphical stimulus are a special case. The test maker is asking the candidate to compare Column A and Column B using a chart, and the answer choices A/B/C/D are the familiar four-line block. Candidates who have prepared for data-interpretation sets sometimes underperform on these because they read the chart the way they would read a set, then try to compute an exact value for each column. That is rarely the right move.

The right move is comparison, not computation. If Column A is 'the median of series X' and Column B is 'the median of series Y,' the candidate should read both medians off the chart and place them against each other. The arithmetic is a comparison, not a calculation. A candidate who starts computing an exact median for one series and an exact median for the other has already spent 90 seconds on a 60-second item.

The other habit that pays off on graphical QCs is the 'no computation' answer. If both columns can be read off the chart and one is clearly larger, the answer is A or B and the candidate moves on. The C ('the two quantities are equal') and D ('the relationship cannot be determined') options exist for items where the chart is ambiguous or where the comparison depends on a missing variable. A senior tutor's advice is to default to the chart, not to a calculation, and to reach for C or D only when the chart genuinely cannot resolve the comparison.

Worked example pattern. A line graph shows 'annual coffee consumption, kg per capita, in country P and country Q, 2014 to 2020.' Column A: 'the increase in country P's consumption from 2014 to 2020.' Column B: 'the increase in country Q's consumption from 2014 to 2020.' The candidate reads the 2014 and 2020 values for both lines, computes the difference for each, and compares. The trap is a candidate who computes the average for one country and the average for the other and then compares those. That is the wrong quantity. A 20-second stem re-read avoids the trap entirely.

Working with two-way tables and frequency distributions

Two-way tables are the most common single-question stimulus in the data-analysis strand, and the most common multi-question stimulus inside the data-interpretation family. They are also the items where arithmetic mistakes cluster, because the candidate is doing two reads — a row and a column — and then a calculation.

The first move on a two-way table is to locate the marginal totals: the row sums, the column sums, and the grand total. Most of the questions in a set are answerable from the marginal totals alone, and a candidate who has read the marginals in 15 seconds will skip the arithmetic for half the questions. A 6-row by 5-column table with marginal totals is a 15-second stimulus and a 60-second per-item average. The same table without marginals is a 30-second stimulus and a 90-second per-item average.

The second move is to identify the percentage question versus the count question. The test maker signals this in the stem: 'What is the ratio of…' is a count question, 'What percentage of…' is a percentage question, and 'What is the probability that a randomly selected…' is a conditional probability question that uses the table as a two-way frequency distribution. Each of these three question types is a different arithmetic pattern, and a candidate who has practised the patterns can switch between them in 10 seconds.

The third move is the cross-row, cross-column read. A question like 'What percentage of the female respondents were in the 30–39 age group?' requires the candidate to read the female row, find the 30–39 column, compute that cell as a fraction of the female total, and convert to a percentage. This is a 30-second read and a 30-second calculation. A candidate who tries to compute it from the grand total has the right number but the wrong denominator. The fix is mechanical: write the denominator beside the cell before you divide.

The arithmetic underneath the chart: which operations actually appear

Data analysis items look like reading questions, but the arithmetic underneath is narrow. Most items require one of five operations: a ratio, a percentage, a difference, an average, or a probability. A candidate who has practised these five operations in the context of graphical stimuli rarely gets lost on test day. A candidate who has practised the operations in isolation, on bare numbers, often stumbles when the operations are wrapped in a chart.

Ratios are the most common. The test asks for the ratio of two bars, two lines, or two cells, and the candidate must read both values and simplify. The trap is a candidate who reads one value from the chart and the other from the stem — the stem often gives a context number, not a chart number, and the two are sometimes the same and sometimes not. A 5-second check of which number came from where kills this trap.

Percentages appear in two forms. The first is the 'percentage of the total' form, which is a single division and a ×100. The second is the 'percentage change' form, which is (new − old) / old × 100. Candidates regularly confuse the two. A 10-second stem re-read, asking 'percentage of what, change from what to what,' resolves the confusion before the calculation starts.

Differences and averages are the cleanest operations. A difference is a subtraction; an average is a sum divided by a count. The test maker plants a trap in the count: a chart with five bars where the test asks for the average of three of them, or a table with one row marked 'n/a' that the candidate must exclude from the average. A 15-second audit of the count before the calculation avoids the trap.

Probability is the rarest of the five but the highest-leverage. The test asks 'What is the probability that a randomly selected respondent is in group X, given that the respondent is in group Y?' The candidate must use a two-way table as a conditional probability distribution: the numerator is the cell at the intersection of X and Y, and the denominator is the marginal total for Y. Candidates who try to compute a joint probability first, then divide, are doing two operations; the conditional form is one operation. The mechanical habit of writing the denominator beside the cell before dividing is the time-saver here.

Preparation strategy: how to train the four-step skim

The four-step skim is a habit, not a fact, and habits need a training plan. The fastest way to internalise the skim is to do 10 to 15 timed data-interpretation sets per week for four to six weeks, each set capped at four minutes, and to score every set on two axes: number correct and total seconds. A candidate who improves on number correct but not on seconds is reading more carefully; a candidate who improves on seconds but not on number correct is reading faster but missing the traps. The target is both, in the same week.

Drill type one: chart-only drills. The candidate is given a single chart and three questions, with a four-minute cap. The goal is to clear the set in three minutes, with one minute of buffer for the synthesis question. After each set, the candidate writes down the trap they almost fell into, in one sentence. Over six weeks, the trap list shrinks from 10 entries to 2 or 3, and the score climbs with it.

Drill type two: table-only drills. The candidate is given a two-way table with marginal totals and a 90-second budget per question. The goal is to clear 8 of 10 in 75 seconds each, with no arithmetic errors. The arithmetic on a two-way table is mechanical, and the only way to get fast is to repeat the pattern until the read and the calculation merge into a single motion.

Drill type three: graphical QC drills. The candidate is given 10 graphical QC items, with a 60-second budget per item. The goal is to recognise the 'no computation' answer and to default to the chart rather than to a calculation. This drill pays off fastest: most candidates gain 5 to 8 points on the Quant section within two weeks of focused graphical-QC work, because the items they were avoiding become the items they are clearing first.

Drill type four: error-log review. Once a week, the candidate opens the error log and groups the missed items by trap type. The categories are usually the same three: rounding, axis misread, and legend confusion. The candidate writes a one-sentence counter-move for each category and re-reads the counter-move before the next practice block. In my experience this is the single highest-leverage study habit in the data-analysis strand: a candidate who reviews 20 missed items per week for six weeks gains more than a candidate who does 40 fresh items in the same week without reviewing.

Scoring implications: where the data-analysis points live

On the GRE General Test, Quantitative Reasoning is scored on a 130 to 170 scale, and the score is adaptive at the section level. That has two practical consequences for data-analysis preparation. The first is that a candidate who cleans up the data-analysis strand first is also lifting the section as a whole, because the test's adaptive logic tends to feed more difficult items to candidates who clear the early ones. The second is that the data-analysis strand is the most score-efficient place to invest preparation time, because the items are reading-heavy and arithmetic-light, so a few weeks of focused work produces a measurable lift.

A 158 on Quant is a common score for a candidate who has practised the arithmetic strands and skipped the data-analysis strand. A 162 to 165 on Quant is the typical outcome for a candidate who has spent the same number of hours on data analysis as on arithmetic. The lift comes from two sources: the candidate is now clearing items they used to skip, and the test's adaptive logic is now feeding them slightly harder items in the second half of the section, where the points are worth more. The lesson is structural, not motivational: in GRE Quant, the strand that looks easiest is the strand that pushes the section score up the most.

For candidates targeting competitive programmes, the data-analysis strand is also the strand that smooths out the worst-case scenario. A candidate who is strong on arithmetic but weak on data analysis will hit a ceiling around 162 to 164, because the second section's harder items lean graphical. A candidate who is balanced across the strands is more likely to break 165, because the test's adaptive engine rewards consistency. The preparation implication is clear: the data-analysis strand is not a side activity, it is a core scorer.

Common pitfalls and how to avoid them: a tactical summary

The data-analysis strand punishes a small number of habits very consistently, and the same habits show up across every candidate population. A senior tutor's job is to name them, drill against them, and turn them into one-line counter-moves the candidate can recite on test day.

1. The rounding trap. The chart shows one decimal; the candidate rounds to a whole number; the answer choices punish the round. Counter-move: keep one decimal in every intermediate calculation, even when the stem says 'approximately.'
2. The axis baseline trap. The y-axis starts above zero; a small bar looks like a tall bar. Counter-move: read the bottom of the y-axis before reading any bar.
3. The legend confusion. A multi-series chart has two colours; the candidate answers for the wrong series. Counter-move: read the legend first, then the stem, then the chart.
4. The skip-question-one trap. The candidate starts at question two to save time. Counter-move: do question one in 45 seconds, use it as a warm-up.
5. The synthesis-overcalculation trap. The candidate computes a value to four significant figures on a chart that supports one. Counter-move: accept the chart's own precision, pick the closest answer choice.
6. The denominator trap on two-way tables. The candidate divides by the grand total when the question is conditional. Counter-move: write the denominator beside the cell before dividing.

Building a six-week plan around the data-analysis strand

A balanced six-week plan treats the data-analysis strand as a primary scorer, not a side activity. Weeks one and two are diagnostic: the candidate takes two full-length Quant sections, scores the data-analysis items separately, and groups the misses by trap. Weeks three and four are drill: the candidate runs the four drill types above, four days a week, with one full-length Quant section on day five. Weeks five and six are consolidation: the candidate reviews the error log once a week, runs timed mixed sets twice a week, and takes one full-length Quant section under realistic conditions.

Within each week, the time budget should split roughly 40 per cent to data analysis, 40 per cent to arithmetic, and 20 per cent to geometry and number properties. This split is unusual — most candidates spend more than 60 per cent of their time on arithmetic — but it matches the section weight and the score-leverage of each strand. A candidate who follows the split for six weeks typically gains 4 to 7 points on Quant, with most of the gain coming from the data-analysis items that were previously skipped.

Day-to-day, the candidate should anchor the data-analysis work to a single drill type per session, with a strict time cap. A 30-minute session of chart-only drills at a four-minute cap per set is more productive than a 90-minute session of mixed drills with no cap. The cap forces the four-step skim; without the cap, the candidate reverts to the slow read that worked in the diagnostic and never builds the test-day rhythm.

Reading a graphical item in 45 seconds: a worked example

A line graph titled 'Average test scores by year, 2015 to 2020, for groups A, B, and C.' Y-axis: score, from 60 to 100, in increments of 5. X-axis: years 2015 through 2020. Three lines, labelled A, B, and C, with A consistently highest, C consistently lowest, and B in the middle. The stem: 'In 2018, the score for group B was approximately what fraction of the score for group A?'

The four-step skim takes 20 seconds. Step one: the title says average test scores, the y-axis is in score units, the x-axis is years. Step two: the legend is the three labels A, B, C; the source line is not present. Step three: in 2018, group A is the highest line, group C is the lowest, group B is in the middle. Group A reads approximately 92, group B reads approximately 82. Step four: no visual trap — the y-axis starts at 60, the increments are uniform, the lines do not cross in 2018.

The arithmetic takes 15 seconds. 82 / 92 is approximately 0.89, which is closest to 8/9 or roughly 0.9. The answer choice is whatever the test offers closest to 0.89. The candidate moves on. Total time: 35 seconds, with a 10-second buffer for a stem re-read. A candidate who had spent 30 seconds on the arithmetic alone would have spent 50 seconds on the item and would not have noticed the absence of a visual trap until the next item.

Conclusion and next steps

GRE Data Analysis rewards a different kind of fluency than the rest of the Quant section. The arithmetic is light, the reading is dense, and the test maker is using the stimulus to test whether the candidate can reframe a problem under time pressure. The candidates who score 165+ on Quant are not the ones who compute fastest; they are the ones who read the chart in 20 seconds, identify the trap in 5, and place the arithmetic in 30. The four-step skim, the 1-2-3 ramp, and the trap list are the three habits that turn the data-analysis strand into a scorer instead of a time sink.

A useful next step is a single diagnostic session: two full-length Quant sections under timed conditions, with the data-analysis items scored and grouped separately. The diagnostic produces a personalised trap list, and the trap list is the input to a six-week plan that follows the split above. TestPrep İstanbul's diagnostic assessment is a natural starting point for candidates building a sharper preparation plan around the data-analysis strand of the GRE General Test.

Frequently asked questions