Two graphs can look dramatically different while displaying exactly the same numbers. Before deciding that one result is larger, faster, steadier or more successful, compare the quantities, units, axes, time intervals and populations. Then calculate the relevant difference or rate. Looking at the picture is a useful beginning; treating its shape as the answer is where the comparison can fail.
This is a practical guide for students reading examination figures, parents comparing progress dashboards and readers checking charts in school projects or public reports. The examples are invented to make the arithmetic inspectable. They are not findings about an actual school, student, product or programme.
One set of numbers, two very different impressions
Imagine a school project weighing material collected for recycling. Three weekly observations are 40 kilograms, 44 kilograms and 48 kilograms. One student draws the line on a vertical scale from zero to sixty. Another draws it on a scale from thirty-eight to fifty.
The second line fills much more of the plotting area. Its rise looks dramatic. The first looks modest. Yet both represent an increase of eight kilograms between the first and third observations. Relative to forty kilograms, that is a twenty percent increase. Redrawing the axis has not collected another gram.
Now make the first chart narrow and tall, and the second wide and shallow. The apparent steepness changes again. A reader who compares the angles without checking the axes is comparing design choices as well as data.
The first useful response is therefore a sentence about quantities: “The recorded mass increased from forty to forty-eight kilograms across these observations.” Only after establishing that sentence should the reader decide what the change means for the project.
Ask what the chart claims before judging whether it works
A graph can serve several legitimate purposes. It may show absolute magnitude, reveal small changes, compare categories, display variation over time or help detect unusual observations. A design suitable for one purpose may be poor for another.
A line chart focused closely on small temperature changes is not automatically dishonest because its vertical axis excludes zero. A bar chart presenting category sizes invites a different reading: the lengths themselves represent magnitude. Cutting away part of those lengths can exaggerate the visual ratio.
The Office for National Statistics guidance on axes and gridlines distinguishes these cases. It recommends zero baselines for bars and areas, allows appropriate cropping for line charts and scatter plots, and recommends consistent scales when comparable charts are placed together. These are useful design safeguards, not a reason to accuse every unfamiliar chart of deception.
For a reader, the central question is whether the visual comparison matches the numerical comparison being claimed. “The change is visible” and “the quantity is enormous” are different statements. A graph should not quietly slide from the first into the second.
The bar-chart trap: a fivefold picture of a twenty percent rise
Return to the invented recycling figures. Suppose bars for forty and forty-eight kilograms begin at a cropped baseline of thirty-eight. The visible lengths above that baseline are two units and ten units. The second visible piece is five times the first.
The actual quantities are not in a five-to-one ratio. Forty-eight divided by forty is 1.2. The later quantity is twenty percent larger, not four hundred percent larger.
This distinction is especially useful when a student says, “The bar is about five times taller, so the amount increased five times.” The error is not necessarily multiplication. The student may have treated a displayed segment as the entire measured quantity.
Repair the interpretation by reading the axis values, stating the original quantities and calculating their relationship. Then explain why the visible bar lengths produced the wrong impression. Merely telling the learner to look more carefully leaves the mechanism of the mistake untouched.
When creating a chart for a school report, use a zero-based bar axis for this magnitude comparison. A labelled difference, a small table or an appropriately designed line chart can make the change visible without misrepresenting the whole quantities.
A slope is a rate only after the units are known
A line rises by eight vertical units across two horizontal units. What is its rate? The answer depends on what those units represent. Eight kilograms over two weeks gives four kilograms per week. Eight kilograms over two days gives four kilograms per day. The drawn angle alone cannot distinguish them.
Even when units match, screen geometry matters. Stretching a chart vertically changes the angle without changing the numerical rate. A line that looks steeper in one panel may not represent a greater rate than a flatter line in another panel.
Teach the learner to calculate change in the measured quantity divided by change in the relevant interval. State both units. Then check that this average rate actually answers the question. An average over a period does not reveal every fluctuation within that period.
In Mathematics and Science, this distinction connects representation to meaning. In reading, it connects a visual to the claim made about it. In public communication, it prevents an attention-grabbing shape from doing the work that a properly specified rate should do.
Equal spaces may hide unequal periods
Consider a second invented record. A quantity is forty units at the start, forty-four after one week and forty-eight after five weeks. A spreadsheet may place the three observations at equally spaced category positions if the horizontal axis is treated as a list rather than a numerical time axis.
The resulting line appears to rise at a steady rate: four units, then four units. But the first increase took one week and the second took four weeks. The corresponding average rates are four units per week and one unit per week.
The value increments are equal; the rates are not. A correct answer must keep both parts of that statement. The fix is to read the dates or elapsed times and, where appropriate, plot them on a genuine time scale.
This also matters when periods are missing. A gap between observations is not automatically a period of no change. Nor does a line joining the available points establish every unobserved value between them. Distinguish what was measured from what the drawing connects.
For subject-specific practice, the guide to reading tables and line graphs with multiple data series provides a route into the underlying reading work. Here the extra task is checking whether two visual presentations support the same comparison.
Counts and percentages answer different questions
Imagine a reading programme in two groups. In Group A, twenty-four of forty students complete a challenge. In Group B, thirty of sixty complete it. The second group has more completions: thirty rather than twenty-four. The first has the higher completion proportion: sixty percent rather than fifty percent.
Both statements are true. A chart of counts and a chart of percentages are not disagreeing. They are answering different questions. The failure occurs when the reader uses one to make a claim that requires the other.
Ask whether the decision concerns total workload, reach or the chance of completion within each group. A teacher arranging certificates may need the count. Someone comparing completion proportions needs both numerator and denominator. Neither chart alone establishes why the groups differ.
Also check whether the group definition changed. “Students invited”, “students who started” and “students who submitted a final response” create different denominators. A rising percentage can reflect a changed denominator as well as a changed numerator.
Write the full statement before reducing it to a label: “Thirty of the sixty students who started completed the task.” That wording keeps the population attached to the number and makes later comparisons easier to audit.
Percentage points are not percent change
Suppose an invented completion rate rises from forty percent to fifty percent. The absolute difference is ten percentage points. Relative to the initial forty percent, the increase is twenty-five percent.
A chart may legitimately emphasise either description, but the accompanying text needs to say which one it uses. “Up ten percent” is ambiguous in this situation. The reader should not have to guess whether the writer means percentage points or relative change.
The baseline matters in the calculation. A later change from fifty percent back to forty percent is a decrease of ten percentage points but a twenty percent relative decrease, because the starting rate is now fifty percent.
These calculations do not reveal the educational significance of the change, its cause or whether the samples are comparable. They repair one narrower problem: translating a visual difference into an accurate numerical sentence.
The chosen time window can reverse the story
Consider five fictional weekly values: thirty, forty, fifty, forty-five and forty. A chart showing only the first three observations rises throughout. A chart showing only the final three falls throughout. A chart showing the first and last records a net increase of ten units.
There is no need to accuse the numbers of inconsistency. The selected window changes the question. “What happened during the recovery?” is not the same as “What happened after the peak?” or “What changed across the full period?”
Before comparing two charts, align their starting and ending points. Ask whether the windows were chosen for a relevant reason, such as a stated programme period, or whether the choice merely makes the desired story look stronger.
Do not assume that the longest available window is always best. An old period may involve different definitions or circumstances. The useful window is the one that fits the question and is reported transparently. Where several views matter, show them and explain what each adds.
Two lines crossing does not prove that one thing caused another
A chart can place study hours and test percentages together. If both rise, a reader may conclude that the extra hours caused the better result. But an accompanying change in topic difficulty, teaching or testing conditions may also matter.
With two different vertical axes, even the visual alignment becomes harder to interpret. The axis ranges can make lines appear to track closely or cross at an interesting point. That crossing does not necessarily correspond to any meaningful equality between hours and marks.
ONS guidance advises avoiding dual-axis charts when they make interpretation misleading or difficult, suggesting separate charts or a single suitable axis with annotation instead. For the reader, the immediate safeguard is to identify which scale belongs to each line before drawing a comparison.
Then separate the description from the causal claim. “These two recorded series rose during the same period” describes the observations. “One produced the other” needs additional evidence. The deeper distinction is explained in How Causal Inference Works.
A non-linear scale is different, not automatically wrong
Some charts use a logarithmic scale. On such a scale, equally spaced marks may represent equal ratios rather than equal numerical differences. For example, values of one, ten and one hundred can be equally spaced on a base-ten logarithmic axis.
A learner reading those gaps as additions of nine each time will misunderstand the representation. The repair is to inspect the labelled values rather than assume that every visual interval represents the same additive change.
A logarithmic display can be appropriate for quantities spanning large ranges or for examining multiplicative growth. It should be clearly labelled and suitable for the audience. It is not a shortcut for making small values look more important.
For a school question, follow the stated scale and syllabus expectations. If the scale type is unfamiliar, ask for explanation before applying a memorised slope rule. Knowing how to calculate does not help when the quantity being calculated has been misidentified.
An index compares change, not the original size
Two library collections begin with different numbers of books. Collection A has one thousand and later has eleven hundred. Collection B has one hundred and later has one hundred and twenty. An indexed chart can set both starting values to one hundred. Their later index values become one hundred and ten and one hundred and twenty.
On that chart, Collection B ends higher because its relative growth is greater. It does not contain more books. Collection A remains much larger and added one hundred books, compared with twenty in Collection B.
The indexed graph is useful when the question concerns proportional growth from a baseline. It is unsuitable as the only evidence for a claim about final collection size. Read the axis label: an index value is not automatically a count, percentage or unit of the original measurement.
When a student compares progress charts that start at a shared reference value, ask what was standardised and what information was removed. A transformation can make one comparison clearer while making another impossible without the original data.
A rising cumulative line can hide a slowing activity
Suppose a fictional class completes ten books in the first week, eight in the second and six in the third. The cumulative totals are ten, eighteen and twenty-four. A chart of that running total rises every week, even though the number completed each week falls.
“The total kept increasing” is correct. “Reading activity accelerated each week” is not supported by these figures. The additions became smaller. To assess the weekly activity, inspect the increments rather than the cumulative level alone.
This distinction appears whenever a dashboard shows total tasks completed, total distance travelled or total items collected. A running total answers how much has accumulated. A per-period measure answers what happened during a particular interval. They can move in apparently different ways without contradiction.
Do not remove the cumulative chart simply because it can be misread. Use it for its proper job, label it clearly and provide period values when readers need to judge pace. In a written answer, name the measure explicitly: “The cumulative total increased, but each week’s addition was smaller.” That one sentence prevents a common confusion between having more and doing more quickly.
A compact audit before comparing two graphs
Begin by naming the measurement in each chart. A title such as “success” is too broad. Is the figure showing students, answers, marks, kilograms, minutes or a proportion? Check the legend and footnotes, not just the headline.
Next identify the population and period. Are the same people, categories and dates represented? If not, decide whether the differences are relevant to the intended comparison and keep them visible in the answer.
Then inspect the representation. Check axis start points, endpoints, intervals and units. Look for breaks, logarithmic scales, missing observations and different scales across panels. Do not assume that identical chart sizes imply identical numerical ranges.
Finally, perform the comparison in words and numbers before accepting the visual impression. Calculate a difference, ratio or rate only when that operation fits the question. If the data needed for the calculation are absent, state that limitation rather than estimating a precise answer from an uncertain picture.
This sequence is deliberately small enough to use during ordinary reading. Its purpose is not to slow every graph into a research project. It is to prevent the first visual impression from becoming the final untested conclusion.
How to write an answer that survives checking
Weak answer: “The second graph shows a much bigger improvement because the line is steeper.” Better answer: “Both graphs show the same increase from forty to forty-eight kilograms. The second uses a narrower vertical range, which makes that increase occupy more of the chart.”
Weak answer: “Group B was more successful because thirty students finished.” Better answer: “Group B had more completions, but Group A had the higher completion proportion: twenty-four out of forty compared with thirty out of sixty. Which measure is appropriate depends on the question.”
Weak answer: “The change was steady because the points rise by the same amount.” Better answer: “The measured values increased by equal amounts, but the time gaps differ. The average rate was lower during the longer interval.”
Notice that stronger answers do not merely include more vocabulary. They connect the observation, numerical relationship and limitation. The sentence becomes a checkable account of what the graph actually establishes.
Repairing a student’s graph comparison
Ask the learner to explain what they looked at first. If the answer is the tallest bar or steepest line, do not immediately supply the correct conclusion. Ask them to read the two relevant values and units aloud.
Then ask what calculation would test the visual impression. If the student calculates correctly once the values are identified, the main problem may be graph reading rather than arithmetic. If the values are read correctly but the ratio or rate is wrong, the numerical operation needs separate attention.
Use a second example with a different source of difficulty. A learner who has just learned to inspect a cropped axis may still miss unequal time intervals. Avoid turning the repair into another single-cue rule such as “all suspicious graphs start above zero”.
The Science guide to diagrams, tables, graphs and evidence connects visual reading with answering scientific questions. The Mathematics Article Directory provides wider routes where calculation or representation needs repair.
When the chart is yours, make disagreement less likely
Retain the underlying table. Label the measured quantity and unit. Choose the display for the comparison you actually intend to make. When placing related charts together, use comparable scales where that supports honest comparison, or clearly explain why scales differ.
Keep the numerical conclusion close to the picture. A short statement such as “eight kilograms more than the first observation” is often more useful than an adjective such as “dramatic”. Where uncertainty or missing observations matter, state them rather than letting the line imply uninterrupted knowledge.
Check the chart on the format in which readers will encounter it. A small phone view may hide footnotes or make scale labels difficult to read. This is an accessibility problem even when the original chart was accurate.
Finally, ask someone else to describe the graph without your explanation. Their interpretation may expose an ambiguity that is invisible to the person who already knows what the chart was intended to mean.
If your project includes a chart copied from another source, retain its attribution and relevant footnotes. Redrawing it does not remove the obligation to explain where the underlying numbers came from.
The test is whether the conclusion survives a redraw
A reliable numerical conclusion should survive changes in colour, aspect ratio and sensible display scale, because it rests on the data and their meaning. If a claim disappears as soon as the chart is drawn differently, investigate whether the original impression was doing too much work.
Graphs are powerful precisely because they make patterns visible quickly. The solution is not to distrust them all. It is to connect that speed to a small set of checks: same quantity, appropriate denominator, clear units, comparable period and correctly interpreted axes.
Look first, but do not stop there. Read the values, name the comparison and test it. The strongest graph reader can explain both what the picture shows and why a different picture of the same numbers might have felt so different.