Table Analysis: reading a data table for exactly what is asked

What a Table Analysis item asks

Table Analysis is one of the five question types GMAC lists for the Data Insights section, alongside Data Sufficiency, Multi-Source Reasoning, Graphics Interpretation and Two-Part Analysis. As the name says, the work is reading a table of data, and this lesson practises the core skill: judging statements against the figures in a table, each statement on its own. GMAC states that for Data Insights questions with several parts, all parts must be correct to score, so one careless row can cost the whole question.

The checks on this page are single-answer multiple choice: they ask how many statements are true, or which combination of statements is true. Every statement still has to be settled on its own before you choose.

Read the frame before the numbers

Before you look at a single figure, read three things:

1. The title and any note under the table. They tell you the period, the population and sometimes a definition ("late" means more than two days after the promised date).
2. The column headings and their units. Thousands or units? Percent or count? Per month or per year? A column headed "Late shipments (%)" is a rate; you cannot add it across rows.
3. What each row is. A row may be a store, a month or a product. Statements that mix rows of different kinds ("the North region") need you to group rows first.

Most wrong answers in Table Analysis come from using the right method on the wrong column.

Match each statement to its columns

Take each statement and ask: which column or columns does it need, and what operation? Statements fall into a few families:

Statement type	What you do
Largest, smallest, rank	Scan or sort one column
"At least", "more than", threshold	Compare each row with the threshold
Median	Sort the column, take the middle value
Mean or total	Add the column (or compare with the median by reasoning)
Ratio between two columns	Divide, or better, multiply the threshold back
Group statements	Add counts for the rows in the group, then compute

Name the family, then do only that work. A statement about one column never needs the other columns.

Sorting is your first tool

A sorted column answers rank, median and "how many exceed" questions at once. When you have seven or nine rows, write the column in order on your scratch paper: it takes seconds and prevents the classic slip of taking the median from the unsorted list. For an odd number of rows, the median is the middle value; for an even number, the mean of the two middle values.

Test a ratio condition by multiplying

"Did the warehouse ship more than 410 orders per staff member?" invites seven divisions. Multiply instead: the condition is $\text{orders} > 0.41 \times \text{staff}$. Multiplying a round threshold is faster and less error-prone than dividing awkward numbers, and it gives a clear yes or no even when the two sides are close.

Combining rows: counts add, rates do not

A rate column (percent late, revenue per employee, growth rate) cannot be summed or simply averaged across rows that have different sizes. To get a group's rate:

$$\text{group rate} = \frac{\text{sum of the counts}}{\text{sum of the bases}}$$

So convert each rate back to a count (rate times base), add the counts, add the bases, and divide once. The plain average of the rates is right only when every row has the same base. The weighted rate always lies between the smallest and largest row rates, pulled towards the rates of the bigger rows.

Compare mean and median without a full calculation

A statement like "the median exceeds the mean" does not always need the mean to the decimal. Find the median, then check whether the deviations of the values from it sum to a positive or a negative number: if the values below the median are further from it in total, the mean is pulled below the median. When the call is close, add the column; the section has an on-screen calculator.

Judge each statement independently

Do not let one statement colour the next. A table that looks "good" for one store may still make a statement about that store false. Settle each statement with its own figures, write a tick or a cross, and only then answer.