Reading a data display means translating its design into quantitative claims. Tables, bars, lines, pictographs, and circle graphs can all show valid data, yet they support different comparisons. The key is to identify the denominator, scale, and meaning of each plotted value.
Establish the reference frame
Read the title, population, time interval, units, and legend. Check whether values are counts, percentages, cumulative totals, or rates. A bar at 15 might mean 15 people, 15%, or 15 thousand dollars.
A table's row and column labels jointly identify an entry. A grouped bar graph needs a legend to distinguish its series. A line graph's horizontal spacing should represent the actual spacing of its time values; do not assume equal-looking segments cover equal time intervals.
Circle graphs are appropriate for mutually exclusive parts of one whole. Their percentages should total 100%, apart from rounding. Overlapping categories require a different interpretation and cannot simply be treated as separate slices of the same total.
Example: percentage share versus count
A school had 400 students last year, with 30% in a club. This year it has 500 students, with 26% in the club. The share fell by 4 percentage points, but membership rose from to
The count increased by 10, or A decrease in share does not imply a decrease in count when the total changes. “Percentage points” describes subtraction of percentages; “percent change” divides a change by its original value.
This distinction is especially important when comparing circle graphs from different years or different-sized populations.
Example: recover a whole from a sector
A circle-graph sector of represents 48 respondents. Since those 48 respondents are one fifth of the total. The total is
A second category with represents of the same total, or 72 respondents. The sector area itself is irrelevant: changing the drawing's radius would not change the proportions.
If a sector percentage has been rounded, an inferred total may be approximate rather than exact. Use the precision stated by the problem rather than inventing extra certainty.
Cumulative data require differences
A cumulative line graph shows 120 tickets sold by Monday and 195 by Wednesday. The number sold after Monday through Wednesday is not 195. A cumulative total cannot decrease unless the definition allows cancellations or revisions.
For noncumulative data, the same plotted numbers might instead describe sales on individual days. The display's label determines which calculation is appropriate. A rising cumulative line can coexist with declining daily sales because the total continues to grow while the increments shrink.
Rates depend on intervals and units
A distance-time graph rises from 30 km at 1 hour to 90 km at 2.5 hours. The average speed over that interval is km per hour. Dividing 90 by 2.5 gives an average from time zero only if the initial position and interpretation justify it.
A line between measured points supports interpolation only under an assumed model of behavior between them. Extrapolating beyond the data is less secure because the pattern may change.
Inspect visual emphasis
Truncated bar axes, inconsistent symbol sizes, and unequal category widths can distort perception. Read exact labels and compare ratios numerically. An association visible in a graph also does not establish that one variable caused the other.
Try it yourself
A company's revenue rises from 80 to 100 million dollars, while its expenses rise from 60 to 75 million. Find the percent change in revenue, expenses, and profit.
Answer: Revenue and expenses both rise 25%. Profit rises from 20 to 25 million, also 25%. Compare changes using each quantity's own original value; the raw increases 20, 15, and 5 are not themselves percentages.
What to remember
Identify what each value counts and which whole or interval it references. Distinguish shares from counts, cumulative totals from increments, and numerical evidence from visual emphasis.