Did you know that a line rising from 120 to 150 can appear to tell a simple success story even when the later number counts more kinds of activity? A beautifully labelled chart may be arithmetically correct and still invite the wrong comparison. The missing clue is sometimes a short note beneath the axis or table.
A chart footnote is a note that qualifies how the displayed data were counted, collected, estimated or compared. Before deciding that one series grew faster than another, check whether the population, time period, definitions, measurement units and data quality are comparable. A 25% numerical increase does not establish a 25% increase in the same real-world activity when one year counts print loans and another includes print plus digital loans.
The 60-second chart comparability check
- Read the title and both axis labels; find what is being measured and the unit.
- Read the source line, footnotes, asterisks and notes beneath the chart.
- Ask whether every series covers the same people, institutions, places and dates.
- Check whether the definition changed: the label might look unchanged while the counted activities differ.
- Check whether figures are actual counts, estimates, samples, indexed values, rates or provisional revisions.
- Only then compute a difference or percentage, and write the limit of the conclusion in ordinary language.
Why a footnote may change the answer
Readers naturally notice bar heights, line slopes and large percentage labels first. A chart’s visual encoding highlights comparison, while the methodological explanation often sits below it in smaller text. But a data series is defined by more than a number. Who or what was counted? For which period? Under what rules? Those details determine whether a comparison answers the question a reader thinks it answers.
The UK Office for National Statistics advises using chart notes and footnotes for essential contextual information and reporting clearly when data are weighted or coverage is incomplete: Chart details: chart text. The underlying idea transfers to school surveys, science tables, public datasets and infographics. A note is part of the evidence, not merely decorative small print.
Two values can have identical numeric units but still count different things. For example, 120 library loans from printed books and 150 loans from print plus e-books are both counts of ‘loans’, but they have different coverage. The key is not whether one number is bigger; it is whether the numbers describe a consistent category.
A hand-checked example: the library chart that seemed to grow by 25%
Illustrative invented data, not real library usage or a report of actual pupil outcomes. Imagine a school newsletter’s bar chart titled ‘Library loans’. It shows 120 in 2024 and 150 in 2025. A note under the chart says: ‘2024: printed books only. 2025: printed books and digital e-books.’ A supplementary table states that 2025 had 132 printed-book loans and 18 e-book loans. Assume for this exercise that both periods cover the same months, school and definition of a printed-book loan.
- A first calculation gives (150 − 120) ÷ 120 × 100% = 30 ÷ 120 × 100% = 25%. That arithmetic is correct as a comparison of the two displayed totals.
- The claim ‘printed-book borrowing increased by 25%’ is not supported, because the 2025 total includes 18 digital loans and the 2024 total does not.
- Compare the common category instead: printed-book loans were 120 in 2024 and 132 in 2025.
- The like-for-like change is (132 − 120) ÷ 120 × 100% = 12 ÷ 120 × 100% = 10%.
- Check totals: 132 print + 18 digital = 150 total in 2025. The 25% change in displayed chart totals and the 10% change in comparable print counts answer different questions.
- We cannot calculate a comparable all-format growth rate from these facts alone, because the 2024 e-book loan count was not supplied.
A careful conclusion would be: ‘In this hypothetical example, recorded print-book loans rose by 10% from 2024 to 2025. The chart’s headline totals rose by 25%, but the scope changed when digital loans were included in 2025; they should not be treated as a like-for-like all-format increase.’ The honest account is more informative, not less.
How to find the exact break in comparability
- Population changed: one year covers one campus, another includes an annex or additional schools.
- Time changed: January–June data are compared with a full calendar year.
- Definition changed: ‘visits’ became ‘visits plus remote logins’, or ‘loan’ now includes digital borrowing.
- Unit changed: thousands of dollars are compared with dollars or actual counts with counts per 100 pupils.
- Measurement method changed: manual tally becomes automated counting, possibly altering detection.
- Estimate status changed: a provisional estimate is compared with a revised or final result.
- Missing values are hidden: an omitted bar may mean suppressed, unavailable or not applicable, not necessarily zero.
Each issue requires a different repair. A changed time window may be standardised if enough underlying monthly data exist. Changed categories may be made comparable when a breakdown is available. A new measurement process may require a cautious note rather than a precise numerical correction. Never invent missing breakdowns merely to manufacture a tidy comparison.
A small diagnostic that reveals misunderstanding
Show a learner only the invented two-bar chart: 2024 = 120, 2025 = 150. Ask what can be computed. They can compute the displayed change of 30 loans and 25%. Then reveal the print-only versus print-plus-digital note. Ask: ‘Does the first calculation now prove printed-book borrowing rose 25%?’ A learner who answers no, identifies the mismatch and requests the 2025 print count has understood what makes statistics interpretable.
If the pupil can calculate 25% but not spot the definitional change, teach the notion of like-for-like measurement. If they spot the changed definition but compute 12 ÷ 150 for the print growth, teach the baseline denominator: compare change in printed loans (12) with original printed loans (120). If they confidently declare 2024 had zero e-book loans, address the difference between an unreported category and a measured zero.
How to write a claim-evidence ledger
Use three short columns: proposed claim, supporting data and caveat. For the invented library example, ‘Displayed totals rose 25%’ has supporting numbers 120 and 150 but carries the caveat ‘definitions changed’. ‘Print loans rose 10%’ uses 120 and 132, with the assumption of matched periods and comparable definitions. ‘Total borrowing grew 25% like-for-like’ must be marked not established, because a comparable 2024 total is missing.
This separates numerical truth from interpretative scope. A good learner can say both ‘the arithmetic is right’ and ‘the conclusion is too strong’. That skill applies to historical charts, school performance dashboards, climate diagrams and surveys, without requiring them to know every specialist methodology in advance.
A practice ladder, not just a warning to read small print
Stage 1 — guided reading
Give a simple table headed ‘School event attendance’. It says 90 in Term 1 and 110 in Term 2. Note: Term 1 records pupils only; Term 2 records pupils and parents. Ask the learner why (110 − 90) ÷ 90 cannot establish pupil-attendance growth. The correct answer is unequal counted groups. Ask what missing number would help: pupils alone in Term 2.
Stage 2 — partially independent calculation
Next, an invented class project records 80 water bottles collected in Week A and 100 in Week B. Note: Week A monitored four school days, Week B monitored five. Assuming each day was measured in the same way, compare the simple average collection rate: 80 ÷ 4 = 20 bottles per day; 100 ÷ 5 = 20 bottles per day. The totals differ, but the per-day averages are equal. The learner must state the daily average is not proof that each individual day’s total was 20.
Stage 3 — independent audit
Present a new chart showing 60 responses to a survey in one year and 75 the next. The note says the first figure covers one class while the second covers two classes. Ask for a two-sentence report that avoids claiming class participation increased. A strong response says the total number of responses increased, but the change in sampling scope prevents a direct statement about either individual class’s response rate.
What students can do across four weeks
- Week 1: Read five chart captions and identify title, units, data source and qualification notes.
- Week 2: Highlight the denominator and identify any changed populations or periods.
- Week 3: Repair two misleading claims by using the actual comparable categories or reporting what cannot be known.
- Week 4: Independently audit a fresh chart and explain one defensible conclusion plus one limitation.
Useful questions for parents and teachers
A parent can ask, ‘What exactly does this bar count?’ before ‘Which bar is tallest?’ If the learner says ‘students’ when the chart counts student-days, return to the unit label. A second effective prompt is ‘What would you need to check before calling this an improvement?’ That moves the conversation from guessing intentions to examining evidence.
Teachers can watch three separate skills: noticing the footnote, deciding whether it changes the comparison, and calculating only after selecting a valid denominator. The student who reads the note but cannot apply it needs transfer practice. The student who identifies the definition mismatch deserves credit even if there is no permissible adjusted percentage to calculate.
Boundary cases: when comparison can still be fair
A methodological change does not automatically destroy every possible comparison. If a source provides a corrected, harmonised series, explain that the numbers have been adjusted and examine its method. When both years supply comparable subcategories, select those subcategories with clear labels, as in the print-loan example. If the note only changes an unrelated display style, it may not affect the relevant quantities. Read what the note says, not just whether an asterisk appears.
Likewise, comparing two charts may be useful for exploring different populations if the intended question explicitly asks about those populations. The mistake occurs when the writer silently treats unlike units, periods or populations as interchangeable.
Frequently asked questions
Is a percentage change always misleading if the footnotes differ?
Not always. First identify whether the footnotes change the meaning of the compared values. Some notes concern unrelated formatting or consistent adjustments.
Should I ignore a chart without footnotes?
No. Read its title, units, axes and source. If important context is missing, describe the limitation and seek the underlying dataset or methodology before making strong claims.
Why does the denominator matter?
A percentage change describes a difference relative to a chosen baseline. For an increase from 120 to 132, the denominator is the starting 120, giving 10%. Using 132 or 150 instead would answer a different question.
What does ‘provisional’ mean?
The number may be revised as more information is collected or checked. State its status and avoid presenting an estimate as the final settled figure.
Can two charts with different vertical scales be compared?
They may be compared numerically if their underlying units, definitions and populations are appropriate. Visual slopes or bar heights alone can mislead when axis ranges differ; read the actual values.
A close worth remembering
The best chart reader is not the fastest person to announce which bar is taller. It is the person who can say what the bars measure, what the notes change, which calculations remain valid and which tempting conclusion is unsupported. A quick footnote check protects the meaning of every later calculation.
Continue with checking whether a citation supports a precise claim, reading a pie chart when sample size is missing, the eduKate Education Hub, and the unchanged Clementi Mathematics reference for diagnosis, guided reasoning and independent checks.