GMAT Table Analysis is the Data Insights item family that looks the friendliest on first contact and punishes the reader who treats the table as background decoration. The question presents a sortable, filterable grid, a short business-style prompt, and three statements, each of which must be judged as true or false, or evaluated as the answer to a single multiple-choice stem. The exam format rewards a candidate who reads the prompt before the grid, who maps the columns mentally, and who lets the table's headers do the work of elimination. Most points lost on this item family are not lost to arithmetic; they are lost to a misread column, a swapped unit, or a statement that becomes true only after a partial filter is applied. This article walks through the structural reading, the column mapping, the common trap shapes, and the pacing that lets a Table Analysis item sit comfortably inside the broader Data Insights section of the GMAT Focus.
What a GMAT Table Analysis prompt actually asks you to do
The first reading pass on a Table Analysis item has a specific job: decide whether the question is a three-statement true/false prompt or a single-answer multiple-choice prompt. The two formats look almost identical at the table level, but the cognitive work they demand is different. A three-statement item is essentially three mini-questions stacked into one grid; a multiple-choice item is one question and the table is the entire source of evidence. In my experience, candidates who skip this classification step end up reading the grid in the wrong order: they skim the table first, then read the prompt, then return to the table, and by the third round the columns have blurred. A clean pass of the stem first sets the search radius, and the table read that follows becomes targeted rather than ambient.
The three-statement format is the more common of the two and is the one most candidates associate with the item family. The stem will say something like 'For each of the following statements, select True if the statement is true and False if the statement is not true,' followed by three labelled statements, often A, B, and C. The candidate's job is to evaluate each statement independently against the table, apply any filters the statement implies, and decide whether the table supports the claim. The three statements are usually calibrated so that two are clearly true or clearly false and the third is the discriminator. A candidate who charges through all three at equal speed loses the section's points on the third. The protocol that protects the score is to flag the third statement as the slow one before opening the grid.
The single-answer multiple-choice format is less frequent but it changes the reading strategy. Here the prompt asks a single question with five answer choices, and the table is the only source of evidence. The choices are typically partial-table reads that diverge by one column, one row, or one aggregation. The trap answers are built by swapping a single variable or by including a row that the filter excludes, and the candidate's job is to read each answer choice against the table with the prompt's filter applied. This format rewards a slower first read of the prompt and a faster, almost mechanical, table check for each answer.
Three behavioural rules hold across both formats. Read the prompt before the grid. Identify the filter words (such as 'in region X', 'for product Y', 'between periods A and B'). Decide whether the prompt is asking a yes/no question, a numeric question, or a categorical question. Once those three decisions are made, the table is a lookup problem rather than a comprehension problem, and the item usually resolves in under three minutes.
The column-first reading protocol for the table itself
Once the stem is locked in, the table read begins with the headers, not the rows. A Table Analysis grid is built from a small number of categorical columns (region, product line, customer segment, period) and one or two quantitative columns (revenue, units, percentage, growth). The sorting and filtering controls exist precisely because the data set is too large to scan row by row. A candidate who reads row by row is doing the work the sorting tool was designed to do. The protocol is: read the column headers, decide which one carries the prompt's filter, and apply that filter mentally before reading any value.
Step one is naming the filter column. If the prompt says 'in the Asia-Pacific region', the candidate must identify the region column and treat every value in that column as the binary check. Step two is naming the metric column. The prompt will ask about a quantity — units sold, revenue, share, growth — and the candidate must identify which column carries that quantity and which carries a related-but-different quantity. The off-by-one column read is the most expensive trap in the item family, and it happens when the candidate picks the column adjacent to the right one because both columns look similar at a glance. Step three is naming the row-count implication. Some prompts ask 'how many segments', others ask 'what is the total'. The first is a count over a filter, the second is a sum over a filter, and the table read is different for each.
For most candidates, I'd personally pick the column-first read over a row-by-row scan because it converts a 200-row grid into a 5-to-10-row problem in about 30 seconds. The trap is that the column headers on a Table Analysis table are often abbreviated or compressed (for example 'Rev Q1', 'Units YTD', 'Penetration %'), and a candidate who skims the headers will mis-apply a filter without noticing. A 10-second pause to read each header literally, expanding the abbreviation in plain English, prevents roughly half of the off-by-one errors I see in diagnostic work.
The last move in the column-first protocol is to confirm the unit. Tables in this item family will sometimes mix absolute numbers and percentages in adjacent columns, or mix thousands and millions, and the candidate who reads '12' without noticing that the column header ends in 'M' will produce an answer that is off by a factor of a thousand. The unit check is the cheapest insurance available on the GMAT Focus and it costs about five seconds. Build it in.
How to triage the three statements in a true/false prompt
The three-statement format is where pacing discipline matters most, and the natural temptation is to treat the three statements as equal-weight tasks. They are not. Statement three is almost always the discriminator, and the exam is engineered so that a candidate who reaches statement three with no time pressure has a good chance of getting it right. The protocol is to do statements one and two fast, accept that they are calibration, and reserve the deeper read for statement three.
A workable split is 60 seconds for statement one, 60 seconds for statement two, and 120 seconds for statement three. This pushes the budget for the item to roughly four minutes, which is consistent with the Data Insights section's overall pacing. Statement one is usually the easiest because it is the most direct read of the table: it asks about a single segment, a single period, or a single value. Statement two adds a comparative element — 'more than', 'less than', 'equal to' — and requires a second column read or a ratio. Statement three typically adds a filter and a computation, and it is the one that separates the 70-percentile from the 80-percentile candidate.
For each statement, the work breaks into four micro-steps. First, translate the statement into a filter plus a check. Second, apply the filter in the table. Third, read the relevant value. Fourth, decide whether the statement's claim matches. If a candidate can complete those four micro-steps in under a minute on statements one and two, statement three inherits the time it deserves. If a candidate gets stuck on statement one, the right move is to mark a tentative answer and move on; the three-statement format is independent, and a wrong answer on statement one does not affect the scoring of statements two and three.
The biggest mis-allocation of time in this item family is the candidate who reads statement three carefully but races statements one and two, then realises on a review pass that statement one was misread because of a column swap. Slow down on the calibration statements; race the discriminator. It feels counter-intuitive, but in my experience the score lift comes from cleaning up the early statements, not from squeezing the last 10 seconds out of the hard one.
Common pitfalls and how to avoid them on a Table Analysis item
Table Analysis rewards a candidate who treats the table as a precise instrument rather than a backdrop. The pitfalls are almost always precision errors, not comprehension errors, and they cluster into four families. Knowing the families in advance changes the way the candidate reads the table, and that is worth more than any single content review.
- Off-by-one column reads. The candidate reads column X when the prompt asks about column Y. Mitigation: name the column out loud or in your head before reading any value. A 5-second check beats a 60-second rework.
- Filter that is not exclusive. The prompt says 'in region X' but the candidate includes a row from region Y because the row's other fields overlap. Mitigation: re-read the filter words before committing to the answer. The filter is the candidate's contract with the table.
- Unit and scale slips. The column is in millions and the candidate reads it in thousands, or the column is a percentage and the candidate treats it as a count. Mitigation: glance at the unit symbol on the header before reading any value.
- Partial-truth statements. Statement three is true for two of the three filters the prompt implies, and the candidate marks it true because two out of three felt close enough. Mitigation: a partial truth is a false. The statement must hold under every filter the prompt names, not most of them.
- Reversed comparatives. The statement says 'A is greater than B' and the candidate reads B as greater than A. Mitigation: write the comparison in plain English before reading the values; the act of writing it forces a check.
A second, subtler pitfall is the candidate who relies on the table's sort order. Table Analysis tables are usually pre-sorted by one column, often the first column, and a candidate who assumes the sort is on the relevant column will read off the wrong row. The sort is a hint, not a guarantee, and the candidate who wants to be safe will re-sort on the filter column before reading. The five-second sort costs less than the minute it takes to recover from a wrong row.
Finally, the candidate who runs out of time on a Table Analysis item is usually the candidate who opened the table first. The table is the largest object on the screen and the eye is drawn to it. Reading the prompt first is unglamorous, and that is exactly why it is the move that separates a clean pass from a frantic one. The prompt is small, the prompt is fast, and the prompt is the part of the item that tells the candidate what to look for. Start there, every time.
Multiple-choice Table Analysis: how the trap answers are built
The single-answer multiple-choice format is structurally simpler but it is built around a specific kind of trap. The five answer choices are almost always partial-table reads that differ from one another by exactly one variable, and the candidate's job is to identify which variable the prompt actually asks about. The trap is rarely arithmetic; it is a swapped column, an extra filter, or a missed row. The candidate who reads each answer choice against the table with the prompt's filter applied will usually eliminate three choices in under a minute and spend the rest of the budget on the final two.
The cleanest way to read a multiple-choice Table Analysis item is to ignore the answer choices on the first pass. Read the prompt, apply the filter in the table, find the relevant rows, and compute the value the prompt asks for. Then read the answer choices and look for the one that matches the computed value exactly. This 'answer-blind' pass prevents the candidate from anchoring on the first plausible choice and missing the trap. The trap is almost always the choice that is almost right: a number that is off by one, a row that is excluded by the filter, or a column that is adjacent to the right one.