The Data Insights section includes Table Analysis and Graphics Interpretation question types. Both test your ability to read, interpret, and calculate from data displays accurately.

A) Data Sets and Terminology

• Data set: an organized collection of data about a specific topic. A data set can be shown with one or more tables, charts, or both.
• Variable: a named column in a simple table; each variable holds one type of data.
• Dependent variable: its values depend on the values of other variables.
  Example: profit = revenue − expense; profit is the dependent variable.
• Independent variable: its values do not depend on the values of other variables in the data set.
• Record: one case in a data set. In a table, a record is a row. In a chart, a record may be a bar or a point.
• Data point: the value of one variable for one case — a single cell in a table.

B) Data Types: What the GMAT Expects You to Know

• Qualitative (nominal) data: non-numerical categories with no meaningful order (city names, product types). Mode applies; mean and range do not.
• Qualitative (ordinal) data: ordered categories where the "distance" between values is not numerically consistent (ratings, ranks, weekdays). Mode and median apply; mean does not.
• Binary data: qualitative data with exactly two values (true/false, yes/no). Tables often show binary data with a check mark or blank.
• Partially ordered data: qualitative data where some but not all cases can be compared (e.g., a family tree — ancestor/descendant relationships within a branch, but cousins on different branches cannot be directly compared).
• Discrete quantitative data: counts whole units — there can be 3 or 4 students but not 3.5. Mean, median, mode, range, and standard deviation all apply.
• Continuous quantitative data: can be infinitely divided (temperature, time, weight). For any two values, there is always a third between them.

C) Interval vs. Ratio Data

These two subtypes of quantitative data differ in what zero means — this determines what calculations are valid.

• Interval data: zero is an arbitrary reference point, not the complete absence of the quantity.
  Example: Celsius and Fahrenheit — 0°C does not mean no heat. Ratios of interval values are NOT meaningful: 60°F is not "twice as warm" as 30°F.
• Ratio data: zero means complete absence. Ratios are meaningful.
  Example: 10 kg is truly twice as heavy as 5 kg. Kelvin is ratio scale; Celsius/Fahrenheit are not.
• Key rule: only use division or ratios to compare two measurements when you have ratio data.

D) Chart Types and What to Look For

• Bar chart: compare values across categories. Stacked bars show composition; grouped bars show side-by-side comparison. Always read which type is shown.
• Histogram: a bar chart for continuous data. Bars touch (no gaps). Shows the shape of a distribution — clustering, skew, and spread.
• Line graph: tracks change over time. Focus on trends, rates of change (steeper slope = faster change), and inflection points. Two lines crossing means their values are equal at that point only — check before and after.
• Scatter plot: shows relationship between two quantitative variables. Questions ask about correlation direction, strength, outliers, or specific data points.
• Bubble chart: a scatterplot where a third quantitative variable is encoded as the size of each bubble. Larger bubbles = higher values of the third variable.
• Pie chart: shows proportions of a whole (must sum to 100%). Combined with a total, you can calculate absolute amounts from percentages. A larger slice means a larger share, not necessarily more in absolute terms if pie totals differ.
• Combination charts: two chart types overlaid with two different y-axes. Read each axis separately — never mix scales.
• Box plots: show median (middle line), quartiles (box edges), and range (whiskers). Outliers appear as isolated points beyond the whiskers.
• Network diagrams: nodes connected by lines. Directed (arrows) mean one-way relationships; bidirectional means mutual. An arrow from A to B does not imply one from B to A.
• Tree diagrams: show partly-ordered data (org charts, family trees, probability branches). You can infer ancestor/descendant relationships within a branch but cannot compare people on different branches unless the diagram provides that data.
• Flowcharts: show a process. Follow arrows through decision nodes (diamond-shaped) and action nodes (rectangular/oval) to terminal nodes. Appear often in MSR tabs.
• Gantt charts: project scheduling. Horizontal axis = time; each row = a task; each bar shows start, duration, and end. Use to identify overlapping tasks, dependencies, and the critical path.

E) Logarithmic Scales

On a logarithmic axis, equal visual distances represent equal multiplying factors, not equal additions. Moving one unit on the axis might mean "10 times as much," not "10 more."

• A "straight line" on a log-scale chart represents exponential growth in reality.
• Read the scale values from axis labels — do not estimate from visual distances.
• Example: decibels use a log scale — 30 dB is 10 times as loud as 20 dB, not 10 units louder.

F) Distributions

A distribution describes how values of a variable are spread across its possible range. Histograms are the primary tool for visualizing distributions.

• Uniform distribution: all values occur roughly equally often — histogram bars are approximately equal height.
• Normal distribution: symmetric bell curve. Most values cluster near the center; mean = median = mode. About 68% of values fall within one standard deviation of the mean; ~95% within two.
• Right-skewed (positive skew): tail extends right. Most values are low, but a few high outliers pull the mean above the median. Mean > median > mode.
• Left-skewed (negative skew): tail extends left. A few very low values pull the mean below the median. Mean < median.
• Key rule: the tail always pulls the mean toward it. The median is more resistant to extreme values than the mean.

G) Standard Deviation and Spread

Standard deviation measures how spread out values are around the mean. A larger SD means more dispersal; a smaller SD means tighter clustering.

• Two distributions can have the same mean but very different standard deviations — one tight and peaked, the other broad and flat.
• On GMAT DI, you may need to compare the spread of two histograms or estimate which group has more variation, even without computing exact numbers.

H) Trends and Correlations

• Positive correlation: as X increases, Y tends to increase. Scatter plot shows upward-sloping cloud.
• Negative correlation: as X increases, Y tends to decrease. Scatter plot shows downward-sloping cloud.
• No correlation: X and Y values are unrelated. Scatter plot shows no discernible pattern.
• Strength of correlation: the closer dots cluster to the trend line, the stronger the correlation.
• Correlation does NOT imply causation — a classic GMAT trap, especially in Graphics Interpretation.
• Trends over time: upward, downward, cyclical, or volatile. A trend observed in the past does not guarantee continuation — use hedged language: "the data suggest," not "the data prove."

I) Associations with Qualitative Variables

• Two nominal variables are associated when values of one tend to appear with specific values of the other (e.g., Brand A preference correlates with urban residence).
• A nominal variable and a quantitative variable are associated when specific values of the nominal variable tend to coincide with higher or lower quantitative values (e.g., a job title associated with a higher salary range).

J) Generalizing from a Sample

• A larger, randomly selected sample provides stronger evidence than a small or self-selected sample.
• Patterns observed in one population or time period may not apply to a different population or later period.
• Correct DI inferences use hedged language: "the data suggest," "based on the observed period," "if conditions remain similar." Absolute language is almost always wrong.

K) Table Analysis

You see a sortable data table and 3–5 statements to evaluate as True/False (or Yes/No). Each statement is evaluated independently.

• Use the sort function strategically: to find max/min values, sort by the relevant column. This is faster than scanning all rows.
• Read column headers and units carefully before answering. The most common errors come from misreading units (thousands vs. millions) or confusing percent vs. percentage point.
• For each statement, identify the relevant column(s) and rows before computing anything.
• Common trap: the title often says numbers are "in thousands" or "in millions" — every cell must be scaled accordingly.

L) Graphics Interpretation

GI questions present a graphic (chart, graph, or diagram) and ask you to complete 1–2 statements using drop-down menus.

• Study the graphic carefully before the statements: note the chart type, what both axes represent, the units, and the scale (especially whether the vertical axis starts at zero).
• Read all choices in each drop-down before selecting: a choice that seems right initially may be contradicted by actual data values.
• Axis labels and legends are essential: many GI errors come from misidentifying which line, bar, or color represents which variable.
• Data you cannot see is data you cannot use: if the graph does not show a value for a particular year or category, you cannot infer it from the graph alone.

M) Key Calculation Types

• Percent of total: (part / total) × 100
• Percent change: (new − old) / old × 100
• Ratio: value₁ / value₂
• Weighted average: Σ(weight × value) / Σ(weights)
• Approximation: most graph questions allow rounding. The answer choices are spread enough that precision beyond 5% is rarely needed.

N) Common Traps

• Truncated y-axis: a y-axis starting at 80 (not 0) makes bar differences appear exaggerated. Always read the axis scale, never eyeball relative heights directly.
• Units mismatch: question asks for millions, chart shows thousands. Convert before computing.
• Absolute vs. relative: a segment growing as a percentage of a pie may be shrinking in absolute terms if the total is falling. Separate these two types of change.
• Correlation vs. causation: scatter plots show correlation; they do not establish causation. Selecting a causal inference from a scatter plot is a trap.
• Interval data ratio trap: temperature in Celsius/Fahrenheit uses interval scale — 60°F is not "twice as hot" as 30°F. Ratios of interval-scale values are not meaningful (unlike ratio-scale data like weight or distance).
• Non-zero baseline: always check whether the y-axis starts at zero. Visual size differences dramatically overstate actual differences when the baseline is truncated.
• Log scale: on a logarithmic axis, equal visual distances represent multiplicative (not additive) differences. A "straight line" on a log scale is exponential growth in reality.

O) Key Business & Finance Formulas for DI

These appear frequently in Graphics Interpretation (financial charts) and Two-Part Analysis (business scenarios):

• Compound interest: A = P(1 + r/n)^nt where P = principal, r = annual rate, n = compounding periods per year, t = years
• Continuously compounded: A = Pe^rt
• Rule of 72: doubling time ≈ 72 / (interest rate %). Quick mental estimate — at 8%, money doubles in ~9 years.
• Profit margin: Profit / Revenue × 100
• Markup: (Selling Price − Cost) / Cost × 100
• Break-even units: Fixed Costs / (Price − Variable Cost per Unit)
• ROI: (Gain − Cost) / Cost × 100
• Expected value: EV = Σ(xᵢ × pᵢ) — multiply each outcome by its probability and sum. Appears in TPA and probability-based GI.

P) DI Timing Benchmarks

20 questions in 45 min — budget aggressively. MSR is a "bundle" investment: one passage serves 3 questions, so careful reading pays off. TA and GI are typically faster to execute; use saved time to offset longer TPA sets.

Q) Calculator Best Practices (DI Only)

The on-screen calculator is available only in Data Insights, not in Quant or Verbal. Use it strategically:

• Use memory functions: M+ stores a value, M− subtracts from memory, MR recalls memory, MC clears memory. Chain operations rather than writing down intermediate results.
• Use it for: compound interest calculations, long division, square roots, percent-of-percent chains, weighted averages with messy numbers.
• Avoid using it for: simple mental math (10% of 400), rough order-of-magnitude estimates, DS questions (sufficiency doesn't require computation).
• DS trap: do not calculate the actual answer on DS questions — you only need to determine if a unique answer exists. Using the calculator on DS wastes time and pulls focus toward solving rather than sufficiency-checking.

R) What the DI Section Tests (Five Skills)

The Data Insights section measures five skills that appear across all question types:

• Apply: use principles or context from the data in a new situation, or determine whether new information fits or breaks a given trend.
• Evaluate: judge whether data supports an argument, plan, or course of action.
• Infer: draw unstated conclusions that logically follow from the data — including what a term means in context, the outcome when applying a formula, or the relationship between two variables.
• Recognize: identify information given explicitly in the data, including details, agreements or disagreements between sources, and relationships.
• Strategize: find ways to work toward a goal given constraints — for example, choosing a combination of options that meets all requirements at the lowest cost.