Data Visualization: Making Numbers Tell Stories

The same numbers, two different feelings

Suppose your school's attendance register shows these percentages for Class 8's five sections this month: 8A is 82%, 8B is 95%, 8C is 78%, 8D is 91%, and 8E is 88%. Read that sentence again. Which section needs a visit from the class teacher? You probably had to re-read the numbers, hold them in your head, and compare them one by one. Now imagine five bars of different heights sitting side by side, with 8C's bar clearly the shortest. You would spot the problem section in under a second, without doing any comparison in your head at all — your eyes would do the comparing for you.

Here is a second example, this time about change over time instead of comparison between groups. Suppose the rainfall recorded in a town during the monsoon months was: June 180 mm, July 320 mm, August 290 mm, September 150 mm. As a list, you can work out that July was the wettest month, but it takes effort — you have to scan back and forth. Plotted as a line rising from June to July and then falling through August to September, the shape of the monsoon season — a build-up, a peak, and a retreat — is visible instantly, as a shape, not as four separate facts you have to hold in memory at once.

This is the entire reason data visualization exists. The numbers in both examples do not change when you draw them as a chart — 82 is still 82, 320 is still 320. What changes is how fast and how accurately a human brain can extract the pattern hiding inside those numbers. A chart is not decoration added after the "real" analysis is done. Choosing the right chart, and drawing it correctly, is itself a form of analysis — arguably the part that decides whether anyone actually understands what your data is telling them.

Why pictures beat lists: position and length

Your visual system is extremely good at judging two things almost instantly, without conscious counting: the position of a point along a line, and the length of a bar. This is not a coincidence of chart design — it reflects how the brain processes vision. Comparing the heights of five bars uses the same fast, automatic visual judgment you use to tell that one classmate is taller than another just by looking. Comparing five percentages typed as text (82%, 95%, 78%, 91%, 88%) instead uses slow, effortful mental arithmetic — you are subtracting numbers in your head.

This is why a bar chart, a line chart, and a pie chart are not interchangeable decorations — each one is built to exploit a different one of your brain's fast visual judgments: bars exploit length, lines exploit the position and slope of a path, and pie slices exploit angle and area. Picking the wrong one for your question is like trying to measure temperature with a ruler — the tool and the question do not match, and the chart ends up either useless or, worse, misleading. We will come back to exactly how to pick correctly. First, every chart — no matter which type — is built from the same small set of parts, and you need to know all of them before you can either read a chart critically or draw one correctly.

Anatomy of a chart: five parts to know

Look at any chart in a newspaper, in your Class 8 Computer Applications textbook, or in a spreadsheet program like Google Sheets or Excel, and you can always find the same five building blocks:

• Title — a short sentence telling you what the chart is about and, often, what time period or group it covers. A chart without a title forces the reader to guess.
• Axes — the two reference lines (usually one horizontal, one vertical) that the data is measured against. The horizontal axis is called the x-axis, the vertical one the y-axis. Each axis also carries a label naming what it measures — "Score range" or "Number of students" — and, on a bar chart, the category names sit along the x-axis as part of it.
• Scale — the numbers printed along an axis, spaced at equal intervals (0, 1, 2, 3 … or 0, 20, 40, 60 …). The scale is what turns a bar's raw length in centimetres into a meaningful value like "4 students."
• Gridlines — the faint horizontal (or vertical) lines that extend the scale's tick marks across the whole plot, so your eye can trace from the top of a bar straight across to the number it represents without a ruler.
• Marks — the actual visual objects that encode the data: bars in a bar chart, the connected points in a line chart, wedges in a pie chart. This is the part that changes shape depending on chart type; the other four stay conceptually the same across all of them.

The diagram below labels all five parts on one chart, so you can see exactly where each one sits.

Choosing the right chart for the question you are asking

Every chart type is a good match for exactly one kind of question. Picking based on "which one looks nicer" is how misleading charts get made.

Bar chart — compares separate categories. Use it when your x-axis holds distinct, unordered-or-nominally-ordered groups: sections 8A–8E, subjects, states, brands. Each bar's length is the only thing that matters — bar charts must therefore start at zero, because length only means something if it is measured from a true zero point. The five sections' attendance percentages from the opening example belong here.

Line chart — shows change over a continuous sequence, almost always time. Use it when the x-axis values have a natural order and equal spacing — days, months, years. The rainfall-by-month example belongs here: June, July, August, September are consecutive, equally-spaced points in time, and what matters is the slope between them — how fast the value is rising or falling — not just each individual height. Never use a line chart to connect categories that have no natural order, like five unrelated sections or five different subjects — a line drawn between "8A" and "8B" implies a smooth journey between them that does not exist.

Pie chart — shows how a fixed whole splits into parts. Use it only when your categories are parts of one total that add up to 100%, and there are few enough slices (roughly six or fewer) that each one is still readable. "How is a 24-hour day split between sleep, school, study, and other activities" is a valid pie chart question because the parts must add up to exactly 24 hours. "Which of five sections has the best attendance" is not a valid pie chart question, because the five percentages do not add up to any meaningful total — 82% + 95% + 78% + 91% + 88% is not "100% of anything."

With the three chart types anchored to three different questions, let's work two full numeric examples the way a program would actually build them — one for the bar chart's underlying data, one for the pie chart's angles.

Worked example 1: turning raw scores into a bar chart (a histogram)

A chart type called a histogram is a bar chart that groups numeric values into equal-width ranges, called bins, and counts how many values fall in each bin. It is the standard way to see how a set of scores is distributed — bunched at the top, bunched at the bottom, or spread evenly. Suppose 20 students score out of 100 on a test:

scores = [56, 78, 45, 89, 92, 67, 73, 81, 59, 64,
          71, 88, 95, 52, 77, 69, 83, 61, 74, 90]

We choose a bin width of 10, starting at 40 (the lowest score is 45) and ending at 100. That gives six bins: 40-49, 50-59, 60-69, 70-79, 80-89, 90-99. Here is the algorithm as a small Python function:

def make_histogram(data, bin_width=10, start=40, end=100):
    num_bins = (end - start) // bin_width       # 6 bins
    counts = [0] * num_bins
    for value in data:
        index = (value - start) // bin_width
        if index == num_bins:                    # value == end exactly
            index -= 1
        counts[index] += 1
    return counts

Trace the logic before running it. num_bins = (100 - 40) // 10 = 6, so counts starts as [0, 0, 0, 0, 0, 0], one slot per bin (index 0 is 40-49, index 1 is 50-59, and so on up to index 5 for 90-99). For each value, index = (value - start) // bin_width finds which bin it belongs to using integer (floor) division. Take the first value, 56: index = (56 - 40) // 10 = 16 // 10 = 1, which correctly points to the 50-59 bin. Take 95: index = (95 - 40) // 10 = 55 // 10 = 5, correctly the 90-99 bin. The safety check if index == num_bins exists because a perfect score of 100 would otherwise compute index = (100-40)//10 = 6, one slot past the end of the array — the check folds a value of exactly 100 back into the last bin instead of crashing.

Running the trace on all 20 values and sorting each into its bin gives: 40-49 gets {45} → 1 value; 50-59 gets {56, 59, 52} → 3 values; 60-69 gets {67, 64, 69, 61} → 4 values; 70-79 gets {78, 73, 71, 77, 74} → 5 values; 80-89 gets {89, 81, 88, 83} → 4 values; 90-99 gets {92, 95, 90} → 3 values. So make_histogram(scores) returns [1, 3, 4, 5, 4, 3]. As a check, these six counts must add up to the original 20 students: 1+3+4+5+4+3 = 20. They do. These six numbers are exactly the bar heights drawn in the chart labeled above — the bars rise from the 40-49 bin up to a peak at 70-79, then fall away again, telling you at a glance that most of the class scored in the 70s, with fewer students at either extreme.

Worked example 2: turning parts of a whole into pie-slice angles

A full circle is 360°. A pie chart converts each category's share of the total into a slice of that circle, proportional to its share. Suppose a student logs how a 24-hour day is spent: Sleep 8 hours, School 6 hours, Study 3 hours, Play/Sports 2 hours, Screen time 2 hours, Other (meals, travel, etc.) 3 hours. Check first that these add up to a full day: 8+6+3+2+2+3 = 24. Good — this data is valid for a pie chart precisely because it sums to one meaningful whole.

The rule for each slice's angle is: angle = (category value ÷ total) × 360°. As a function that also tracks where each slice starts and ends around the circle:

def pie_angles(values):
    total = sum(values)
    angles = []
    start_angle = 0
    for v in values:
        sweep = (v / total) * 360
        angles.append((start_angle, start_angle + sweep))
        start_angle += sweep
    return angles

Trace it with values = [8, 6, 3, 2, 2, 3], so total = 24. Sleep: sweep = (8/24) × 360 = 0.3333... × 360 = 120°, occupying 0° to 120°. School: sweep = (6/24) × 360 = 90°, occupying 120° to 210°. Study: sweep = (3/24) × 360 = 45°, occupying 210° to 255°. Play: sweep = (2/24) × 360 = 30°, occupying 255° to 285°. Screen time: another 30°, occupying 285° to 315°. Other: sweep = (3/24) × 360 = 45°, occupying 315° to 360°, exactly closing the circle. As a check, the six sweeps must add up to 360°: 120+90+45+30+30+45 = 360. They do — this is the same kind of check you used for the histogram counts, and it is worth doing every time, because a coding mistake in the loop (for example, forgetting to update start_angle) would make the slices overlap or leave a gap, and the angle check would catch it immediately.

Common misconception: an axis that doesn't start at zero is "just zooming in"

A bar's entire visual meaning comes from its length, measured from the axis baseline. If that baseline is not zero, the length no longer represents the value proportionally — and the bar lies, even though the number printed above it is accurate. This is one of the most common ways charts mislead people, often without any dishonest intent, simply by "zooming in for detail."

Here is the pixel arithmetic that proves it. Suppose Student A scores 72 and Student B scores 76 — a real difference of 4 marks out of 100, about 5.6% more. We draw both as bars in a plot area 200 pixels tall.

Full, honest scale (0 to 100): the axis covers 100 units across 200 pixels, so each unit of score equals 200 ÷ 100 = 2 pixels. Bar A's height = 72 × 2 = 144 px. Bar B's height = 76 × 2 = 152 px. The two bars differ by only 8 px out of 200 — visually, they look almost the same height, which is honest, because the students' scores are, in fact, almost the same.

Truncated scale (70 to 80): the axis now covers only 10 units across the same 200 pixels, so each unit of score equals 200 ÷ 10 = 20 pixels — ten times more sensitive. Bar A's height = (72 − 70) × 20 = 40 px. Bar B's height = (76 − 70) × 20 = 120 px. Now Bar B is three times as tall as Bar A (120 ÷ 40 = 3), even though the underlying difference is still the same 4 marks out of 100. A reader glancing at the truncated chart would conclude Student B massively outperformed Student A, when in reality both scored in the same narrow band. The diagram below draws both versions from the exact numbers above, side by side.

The rule that follows is simple and worth memorizing: a bar chart's y-axis must start at zero, because a bar's length is only a fair comparison when it is measured from nothing. Line charts are a partial exception — because a line encodes a trend through position and slope rather than a bar's length from zero, a line chart's y-axis can sometimes be zoomed in to show a trend more clearly, but only if the axis is clearly labeled with its true starting value so no reader is misled into assuming a hidden zero.

Reading the anatomy and the chart-choice rule together

You now have both halves of the skill. The anatomy section tells you how to read any chart's title, axes, scale, gridlines, and marks without confusion. The chart-choice section tells you which of the three basic chart types actually fits your data before you even start drawing — categories get bars, a time sequence gets a line, and only a set of parts that sum to one true whole gets a pie. This is exactly the pairing your Class 8 Computer Applications and Artificial Intelligence curriculum is building toward: constructing and correctly reading charts from spreadsheet data, a skill that then gets extended with formulas and larger datasets in Class 9-10 Computer Applications, and formalized further in the Informatics Practices elective some of you may choose in Class 11-12. Whichever path you take, the two checks used throughout this chapter — do the bin counts add up to the total number of data points, and do the pie angles add up to 360° — are exactly the kind of self-verification that class tests, projects, and real-world data analysis actually reward, because they catch your own mistakes before anyone else has to point them out.

Active recall

• A class has 5 sections with attendance 82%, 95%, 78%, 91%, 88%. Which chart type should you use to compare them, and specifically why would a line chart be the wrong choice here?
• Rainfall by month was recorded as June 180 mm, July 320 mm, August 290 mm, September 150 mm. Which chart type fits this data, and what does its shape reveal about the monsoon that the raw list does not show as quickly?
• Using the make_histogram logic with bin width 10 starting at 0, which bin does a score of 67 fall into? Show the index calculation, the way the chapter did for 56 and 95.
• A pie chart represents a 24-hour day. Screen time is 3 hours. What angle, in degrees, should that slice occupy? Write out the formula and the arithmetic.
• A bar chart's y-axis is truncated to start at 50 instead of 0. Two bars represent scores of 82 and 84. Will the visual height difference look bigger or smaller than the true 2-mark difference? Explain using the pixel-height reasoning from this chapter, not just a guess.

Check your work: (1) a bar chart — the five sections are unordered categories, not points in a continuous sequence, so a line between them would falsely imply a smooth trend connecting unrelated groups. (2) a line chart — the rise from June to July and the fall from July to September form a peak shape, showing the monsoon builds up and then recedes, a pattern much harder to see scanning four separate numbers. (3) index = (67 − 0) ÷ 10 = 6 using floor division, placing 67 in the seventh bin (index 6), which covers 60-69. (4) angle = (3 ÷ 24) × 360° = 45°. (5) Bigger — truncating the baseline from 0 to 50 shrinks the axis range from 100 units to 50 units, doubling the pixels-per-unit, which stretches the true 2-mark gap into a visually larger gap than it represents.

Summary

A chart's job is to let a reader's visual system do comparison work that raw numbers force the brain to do slowly and consciously — bars exploit length, lines exploit position and slope over a sequence, pie slices exploit angle within one true whole. Every chart, regardless of type, is built from five parts: title, axes, scale, gridlines, and marks. Choosing a chart type is a matter of matching it to the question — categories compared against each other need bars, a value changing across an ordered sequence like time needs a line, and only parts that sum to one complete whole belong in a pie. Building the underlying numbers is itself an algorithm: a histogram bins raw values with (value − start) // bin_width and self-checks by summing counts back to the total; a pie chart converts each part into (value ÷ total) × 360° and self-checks by summing angles back to 360°. And a chart's honesty depends on where its axis starts — a bar chart's y-axis must begin at zero, because a bar communicates its value through length measured from nothing, and truncating that baseline can turn a 5.6% difference into a bar that looks three times as tall.