GMAT Focus Multi-Tab Reasoning is the item family in the Data Insights section that imitates how a working analyst actually handles information spread across separate views of the same dataset. A candidate is given what looks like a small spreadsheet application, with three or four tabs sitting across the top of the screen, and is asked to switch between them to answer two or three linked prompts. The question is not really a reading test, nor is it a quantitative test in the conventional sense; it is a study in tab selection, prompt interpretation, and disciplined use of the on-screen calculator. Treating the item as a faster version of a textbook question is the single most common mistake, and it shows up in the scaled score more than candidates realise.
What a multi-tab set actually looks like on screen
The interface hands the candidate a small workbook with three to four tabs, each tab showing a clean table of figures rather than a chart. The numbers themselves are usually simple: integer counts, percentages, dates, two-letter country codes, or category labels. The difficulty sits in two places, and only two: which tab actually holds the answer, and which row inside that tab matches the wording of the prompt. There is no scrolling through dense prose, no graph to decode, and no long story to digest. Most candidates finish the on-screen reading of a set in well under two minutes; the question is whether those two minutes were spent on the right tab.
Because the workbook is interactive, the candidate is in control of the order in which the tabs are opened. A candidate who clicks every tab in turn before reading a single prompt will, in practice, start to confuse which column belongs to which table, especially when the columns carry similar headers such as 'Revenue', 'Cost', and 'Profit' on more than one tab. A candidate who reads all the prompts first and then opens a single tab in response to one specific prompt is closer to how the test-makers want the question to be answered, but risks missing the link between the two prompts when the second prompt requires information that was first revealed by answering the first.
The mechanics of the question type also include a calculator icon that sits alongside the prompt, not next to the workbook. The calculator is positioned with the prompt on purpose, so that the candidate treats the arithmetic as something done on the answer, not on the table. The workbook is read; the prompt is computed. Conflating those two actions is the second most common mistake, after tab-order confusion, and it costs the same number of points.
How the prompts are constructed around a single underlying dataset
Multi-Tab Reasoning sets are built so that the same underlying data sits in slightly different cuts across the tabs. One tab might break revenue down by product, another by region, and a third by month. The same row labels appear in more than one tab, but the column headers change. The two or three prompts are written to force the candidate to combine those cuts: 'In region X, which product's revenue grew between Q1 and Q2?' requires reading row 'X' on the regional tab and then opening the monthly tab to confirm the order of the months. A candidate who tries to answer the question from the regional tab alone will see a 'revenue' column that says nothing about quarter-on-quarter movement.
Three prompt shapes dominate the family. The first is the cross-tab verification prompt, which asks whether a figure that appears on one tab is consistent with a figure on another. These prompts are usually phrased as a yes/no question with a one-line justification, and they reward candidates who have noticed that the second tab is essentially a re-cut of the first. The second shape is the filtered aggregate prompt, which asks the candidate to apply a filter (a region, a date range, a product category) and then sum, average, or count across the remaining rows. The third shape is the change-detection prompt, which asks for the difference, ratio, or percentage change between two values, where the two values live on different tabs.
Identifying which of the three shapes is in front of you is the first decision a candidate should make, and it should be made before the first tab is opened. In my experience, candidates who spend 20 to 30 seconds tagging each prompt with one of those three labels save a full minute on the back end, because they know in advance which tab to open first and which column to read. The prompts are short, typically one to two sentences, and re-reading them after the table is in front of you is a worse use of time than reading them carefully once, in isolation.
Why 'read prompts first' beats 'open tabs first'
The instinct on seeing a workbook is to click. The instinct is wrong. Open the prompt pane, read both prompts end to end, and underline, mentally or on the scratch pad, the row label, the column header, and the operation word. The row label is what you will search for on the tab. The column header is the cell you will read once you find the row. The operation word is what you will do with that cell: compare, sum, average, take a percentage of, or take a difference against. Once those three slots are filled, the workbook becomes a lookup exercise, and lookups are fast.
This is also the moment to decide whether the two prompts are independent or chained. A chained prompt uses the answer to the first prompt as the input to the second. Independent prompts can be answered in either order. The default ordering of prompts on screen is not always the optimal ordering, and the candidate is free to solve whichever prompt is easier first. A 30-second decision on order, made at the prompt-reading stage, prevents the 90-second scramble that happens when the second prompt turns out to be easier but is left until the end of a 4-minute tab tour.
The five tab-ordering rules that govern almost every set
Rule one: open the tab whose row label is mentioned in the prompt. If a prompt says 'in region X', open the tab that has 'region' as a row label, not the tab that has 'region' as a column header. Row labels are searchable; column headers are not. Rule two: when two tabs both have the row label, choose the tab whose column header matches the second slot in the prompt. If the prompt asks for revenue, open the tab whose 'Revenue' column is the one you actually need (gross, net, per-unit, total), not the tab where 'Revenue' appears as a secondary column.
Rule three: when the prompt asks for a change between two periods, open the period tab second, not first. The period tab answers the question, but only after the row label has been confirmed on the cut that does not change with time. Rule four: when a prompt uses a threshold language such as 'greater than', 'at least', or 'top three', open the tab that allows sorting or filtering directly. If the on-screen interface exposes a filter or sort affordance on one tab and not on the others, that tab is the one to open. Rule five: when the two prompts share no row label, treat the set as two independent questions and budget time as such; do not let an answer from prompt one contaminate the search for prompt two.
These five rules are not a memorisation trick. They describe how the test-makers build the questions. Every multi-tab set is constructed so that one rule, and usually only one rule, points to the right starting tab. A candidate who internalises the rules stops clicking randomly and starts clicking predictably, and predictable clicking is what keeps a 45-second tab open down to 15 seconds.
Reading a multi-tab table the way an analyst would
Analysts do not read tables top to bottom. They read tables in three deliberate passes. The first pass is the column-header pass: scan every header across the top of the table and decide which column is the one named in the prompt. The second pass is the row-label pass: scan the leftmost column (or whichever column the data designer has used as the key) for the value named in the prompt. The third pass is the cell read: at the intersection, read the value, including its units, and write the value into the answer slot. The three-pass method is faster than reading the whole table, and it is more accurate, because the candidate never engages with the columns that the question is not about.
Two reading traps catch candidates at the cell-read stage. The first is the unit trap: a 'Revenue' column expressed in millions on one tab and in thousands on another. The second is the sign trap: a 'Change' column where positive numbers mean growth on one tab and decline on another. Both traps are visible in the column header, not in the cell, which is why the column-header pass is the pass that has to be done first and done slowly. A candidate who reads the cell before the header is the candidate who answers 2.4 when the question expected 2,400,000.
A third, subtler trap is the rounding trap. Some cells carry decimals to one place, others to two, and a few are integers. When a prompt asks for a percentage, a candidate who copies '0.4' from one row and '12.6' from another without checking the column header's implied precision will produce an answer that is off by an order of magnitude. The fix is the same: column header first, cell second, calculator third.
Arithmetic and the on-screen calculator: when to use it, when not to
The calculator icon sits next to the prompt for a reason. It is built for the operation word identified in the prompt-reading step, not for re-deriving values from the table. The most common over-use is typing a single cell value into the calculator and then re-typing it into the answer field. The keyboard accepts numeric input directly into the answer box; the calculator is unnecessary for a lookup that does not involve a second number.
The calculator earns its place when the operation is one of three things: a sum across more than two rows, a percentage change between two cells, or a ratio between two cells that share a denominator. In all three cases, the candidate should type the two values, perform the operation, and copy the displayed result into the answer field. Rounding inside the calculator is acceptable because the answer field accepts the same precision the test-maker has rounded to. Trying to keep extra decimals in the head is a slower path to the same answer and the more common source of small numerical drift.