The GMAT Focus Data Insights section is the youngest pillar of the revised GMAT, and it is also the section that confuses candidates who arrive expecting a numerical obstacle course. In reality, the section is a reasoning section wearing a quantitative mask. Twenty items are delivered in 45 minutes, drawn from five distinct item families, and the score band runs from 60 to 90 in single-point increments. Every question asks the candidate to read a stimulus, isolate a small number of facts, and execute a targeted judgement. The arithmetic involved is deliberately modest. The lift comes from the reasoning layer on top of it.
This article walks through the five item families one at a time, names the score report metrics worth monitoring, and shows how a coherent preparation strategy differs from a topic-by-topic slog. The aim is to give the reader a working map of the section before they ever open a question bank.
The five item families and what each one is really testing
The Data Insights syllabus is narrower than it first appears. Five item families cover the entire section, and a strong tutor will teach them as five distinct reasoning tasks rather than five data formats. The table below summarises the family, the underlying cognitive demand, and the surface skill the candidate is most likely to over-train.
| Item family | Underlying cognitive demand | Surface skill candidates over-train |
|---|---|---|
| Data Sufficiency | Decide whether given facts are enough to answer a stated question | Solving the underlying math when not needed |
| Table Analysis | Sort, filter, and aggregate a sortable spreadsheet view | Memorising every cell before reading the prompt |
| Graphics Interpretation | Read a chart, then choose a stat that matches a stated property | Eyeballing values rather than working with the printed scale |
| Multi-Source Reasoning | Triangulate facts across two to three linked tabs to judge a claim | Reading every tab in full before forming a hypothesis |
| Two-Part Analysis | Make a single decision with two correct values and three distractors | Solving both parts sequentially instead of jointly |
For most candidates, the family that moves a score the most is the one they are least willing to practise. A tutor who notices a candidate treating Data Sufficiency as a quick yes-or-no routine usually has a candidate who can climb a band within a few weeks, because the work is targeted. By contrast, a candidate who is fluent in table analysis but stumbles on multi-source reasoning often loses 8 to 12 scaled points to a skill gap that two focused sessions would close.
Reading the family as a reasoning task
The cleanest way to internalise the five families is to ask, before reading any data, what judgement the question is asking for. Data Sufficiency asks whether you can answer. Table Analysis asks how the table behaves under a sort or filter. Graphics Interpretation asks which statistic matches a named property. Multi-Source Reasoning asks whether a claim is supported, weakened, or unaffected. Two-Part Analysis asks which pair of values is jointly correct. The candidate who learns to name the judgement out loud before opening the stimulus saves roughly 15 to 25 seconds per item, and over 20 items that adds up to a measurable pacing cushion.
In my experience, candidates who skip this step and dive into the chart first are the ones who run out of time on item 17 and rush the final three. Pacing is rarely about raw reading speed; it is about not reading the same stimulus twice.
How the GMAT Focus Data Insights score band is actually built
The score band for Data Insights runs from 60 to 90, reported as a single integer. A score of 60 is the floor; 90 is the ceiling. Most competitive MBA programmes describe the section as scored in isolation, but admissions committees routinely read it against the Quantitative and Verbal bands to gauge the candidate's reasoning profile. A candidate scoring 78 Data Insights with 76 Quant and 80 Verbal is sending a very different signal from a candidate scoring 84 Quant, 78 Verbal, and 70 Data Insights, even though the headline GMAT score (the sum of the three sections, running 205 to 805) might differ by only a handful of points.
The adaptive engine that builds the section is item-level, not passage-level. The candidate's response to each item updates a running estimate of ability, and the next item is selected to maximise information. There is no separate 'hard module' gate; the difficulty adjusts continuously. The practical consequence is that a weak early streak is recoverable within the section, but a strong early streak does not let the candidate coast. Candidates who treat the first five items as a warm-up are giving up the information advantage the engine is trying to give them.
Score report metrics worth monitoring at the end of a practice test are not the percentile alone. The enhanced score report, which is available free of charge on the candidate portal, splits the section into sub-skill indicators and shows the time spent per item relative to the section median. A candidate whose sub-skill indicators are roughly even but whose timing chart shows long tails on multi-source items has a different preparation problem from a candidate whose sub-skills are uneven but whose timing is consistent. Reading the report as a single number flattens the diagnostic signal.
What the percentile column does and does not tell you
Percentiles compare the candidate's score against a recent reference population. A score of 78 has sat in roughly the 75th percentile band for several admissions cycles, and 84 has hovered near the 90th percentile. The percentile is a useful external benchmark and a poor internal target. Working backwards from a desired percentile ignores the fact that the GMAT is scored as a sum of three sections with diminishing returns at the top: a candidate who is already at 84 Data Insights gains far more by lifting Verbal from 78 to 82 than by pushing Data Insights from 84 to 86. Candidates often misread the percentile column as a target in itself rather than as a translation of a band that should be evaluated alongside the other two.
The preparation strategy that lifts a band within a few weeks
Most candidates who plateau below 75 on Data Insights are practising too many items and reflecting on too few. The single highest-leverage change is to build a short, deliberate review loop around ten items per session rather than to grind through fifty. Ten items, reviewed in writing, take about 70 to 90 minutes including the review; fifty items, reviewed casually, take about the same wall-clock time and produce a fraction of the learning.
The loop has three steps. First, the candidate solves the item under timed conditions and notes only the time spent and the chosen answer. Second, the candidate reopens every wrong item and every item that took longer than 130 seconds, and writes a one-sentence diagnosis: which family, which judgement, which step broke. Third, the candidate writes a single-sentence rule they will apply to the next ten items of that family. Over four sessions the rule list grows long enough to function as a personal error log, and the candidate can see their own pattern of slips far more clearly than any score report can show it.
A second tactic that consistently moves scores is to practise the first five items of every set under stricter time pressure than the rest of the section. The first five items of a 45-minute section absorb about 12 minutes in the median candidate's hand. Cutting that to 10 minutes and using the saved time on the multi-source items near the end typically adds 4 to 6 scaled points over a four-week block, because the engine has already given the candidate an early signal of strong performance and continues to feed harder items, which the candidate now has time to read.
Common pitfalls and how to avoid them
- Treating Data Sufficiency as a math shortcut. The five answer choices are about sufficiency, not about a single correct value. Solving the underlying problem is usually wasted motion.
- Sorting the table before reading the prompt. The table is sortable but the candidate should always read the prompt first, identify the relevant columns, then sort. Sorting first trains the eye to scan and the brain to lag.
- Eyeballing a chart when the printed scale is available. Graphics interpretation items always provide a scale; reading it is faster and more reliable than estimating.
- Reading every tab in a multi-source item. Two of the three tabs are usually red herrings. Read the question stem, identify the claim, then open only the tab that holds the relevant fact.
- Letting a two-part item become two separate problems. The two answers are jointly correct; solving for one and matching the other invites a trap on the distractor pair.
Item-by-item tactics for the five families
Once the family is identified, the candidate should have a small set of habits ready. The habits below are not exhaustive, but each one has produced measurable score lifts in the candidates I have tutored through the section.
Data Sufficiency habits
For Data Sufficiency, the habit is to test statement (1) alone, then statement (2) alone, then both together, in that order, and never to evaluate them as a single block. The answer choice follows from the structure of the testing, not from the value of the underlying answer. A candidate who finds themselves arguing with the arithmetic is on the wrong path; the question is whether the data permits an answer at all. When in doubt on a value-based Data Sufficiency item, the candidate should ask whether the two statements produce a single value or a range, and pick the choice that reflects the range.
Table Analysis habits
For Table Analysis, the habit is to read the prompt, then identify the column to sort and the column to read. The candidate should sort once, read the relevant cell, and answer. Re-sorting for a second sub-question is fine; re-reading every column is not. A common trap is to over-rely on the visible rows: many items ask for an aggregate across filtered rows, and the candidate must mentally hold the filter rule while reading. Writing the filter on the scratchpad, in shorthand, is the most reliable defence.
Graphics Interpretation habits
For Graphics Interpretation, the habit is to start with the two answer options, not the chart. Each option names a property — slope, ratio, increase, decrease — and the candidate should ask which property the chart can support. If neither property maps cleanly to a printed axis label, the candidate is probably looking at the wrong chart or the wrong tab. This single habit eliminates a striking number of the 30-second temptations to choose the more 'interesting' option.
Multi-Source Reasoning habits
For Multi-Source Reasoning, the habit is to read the claim, not the stimulus. Two of the three tabs are decorative; the third usually contains the only fact the candidate needs. Once the claim is in mind, the candidate opens the relevant tab, extracts one fact, and decides. A candidate who finds themselves comparing two tabs in detail is almost certainly solving a problem the question did not ask. Time pressure on multi-source items is structural; the section is engineered so that thorough reading is impossible, and the candidate who accepts that constraint performs better than the one who fights it.
Two-Part Analysis habits
For Two-Part Analysis, the habit is to solve the two parts jointly. Many candidates solve the first part, find a single value, then look for that value among the second part's options. The structure of the item is the other way around: the candidate picks one value from the column and one value from the row, and only one cell in the 5x5 grid is correct. The fastest path is to express both parts as a single calculation, then test the candidate's first guess against the available pairs.
Pacing the 45-minute section without panic
Pacing on Data Insights is not a single number. It is a budget that shifts as the section progresses. Items 1 to 5 are the engine's first signal; spending 100 to 110 seconds each is reasonable and gives the candidate time to read carefully. Items 6 to 12 are the consolidation phase; the candidate should be averaging 100 seconds or less. Items 13 to 20 are the closing phase, and the candidate should be averaging 90 seconds or less, banking 20 to 30 seconds per item to cover the inevitable slow read on a multi-source item.
This phasing matters because the adaptive engine feeds harder items as the candidate performs well, and harder items take longer to read. A candidate who budgets 120 seconds for items 1 to 5 and 90 seconds for items 16 to 20 is, in effect, asking the engine to give them easier items later so the timing works out. The opposite works better: spend less time early, accept harder items, and use the saved time to read them carefully.
For most candidates the real pacing failure is not running out of time on item 20; it is running out of time on item 16 and answering items 17 to 20 in a fog. Building the muscle to recognise the moment a single item is eating 150 seconds, and the discipline to mark it, move on, and return, is the most under-trained skill in the section. The section permits marking, and a marked item is far less costly than a guessed item that saps the remaining energy.
Building a four-week study plan that maps to the section
A workable four-week plan allocates roughly 8 to 10 hours per week, split into two 90-minute sessions and one 60-minute review. The first week is diagnostic: the candidate takes a single timed section, scores it, and writes the personal error log described earlier. The second week is family-by-family, with the candidate spending two sessions on each of two families. The third week is timed sets of ten, with the focus on the personal error log. The fourth week is full-length practice under realistic conditions, including the timed mini-section starting from item 1 with no warm-up.
The plan above assumes the candidate has at least four clear weeks before the test date. A candidate with two weeks should compress the diagnostic into a single session, skip the family-by-family phase, and spend the second week entirely on timed sets. A candidate with eight weeks can add a fifth week of pure error-log review, which usually adds another 2 to 4 points beyond what a four-week plan would have produced.
For most candidates the largest single improvement arrives in week two, when the personal error log first reveals a pattern. The pattern is usually one of three: misidentifying the family, spending too long on a sub-step, or guessing under time pressure. Each of these has a targeted remedy that the candidate can apply within the same week, and the score lift follows within two to three timed sets.
Admissions context: how Data Insights is read by committees
Data Insights is the section that admissions committees most often describe as a tie-breaker. The Quantitative and Verbal bands are the two pillars, and Data Insights is the third signal that the candidate can read a chart, judge a claim, and act on incomplete information. A balanced profile across all three sections reads as a generalist; a strong Data Insights with weaker Verbal reads as a quant-leaning candidate whose reasoning under data ambiguity is a particular strength.
Most top programmes will not state a minimum Data Insights score, but the de facto threshold sits in the mid-70s for the most competitive pools. Candidates applying to programmes that use the section as a screening tool should target the mid-80s to be safe, recognising that the score is one input among several. A 78 with a strong application narrative is a different conversation from a 78 with a weak one, and a candidate worried about a single section should spend some of their preparation energy on the application materials that surround the score.
Conclusion and next steps for the Data Insights candidate
Data Insights rewards candidates who treat it as a reasoning section with five well-defined item families, who build a personal error log instead of a question count, and who pace the first five items strictly so the last five have time to breathe. A four-week plan built on those three habits is enough to move most candidates up at least one band, and a candidate who is already in the mid-70s can usually push into the low-80s within the same window. The single highest-leverage next step is to take a single timed section, write the personal error log honestly, and let the log dictate the next ten items. TestPrep İstanbul's diagnostic assessment on the Data Insights item families is a natural starting point for candidates building that sharper preparation plan.
Frequently asked questions
How many questions are on the GMAT Focus Data Insights section?
The section contains 20 questions delivered in 45 minutes. The questions are drawn from five item families, and the difficulty adjusts item by item based on the candidate's running performance.
What is a good Data Insights score for top MBA programmes?
A score in the mid-70s keeps most candidates in the conversation, and a score in the mid-80s signals strong reasoning under data ambiguity. The exact target depends on the candidate's Quantitative and Verbal bands and on the programme's overall applicant pool.
Is Data Sufficiency part of Data Insights?
Yes. Data Sufficiency is one of the five item families on Data Insights, and it is the family that rewards structured testing of the two statements rather than solving the underlying problem.
How should I split my study time across the five item families?
Spend the first week on a diagnostic, then weight the next two weeks towards whichever two families produced the most entries in your personal error log. The remaining week is for timed sets that mix all five families.
Does the adaptive engine change after a wrong answer?
Yes. The engine re-estimates the candidate's ability after every response and selects the next item accordingly. A weak streak is recoverable within the section, and a strong streak leads to harder items that the candidate now has time to read carefully if pacing is well managed.