Blog›Data Analytics Basics›Chapter 10
BASICS · CHAPTER 10Beginner–Intermediate

Data Visualisation — Choosing the Right Chart

Bar, line, scatter, pie, heatmap, box plot, histogram — when to use each one, when not to, the rules that make charts clear instead of confusing, and Indian business examples for every chart type.

📊 Bar Chart📉 Horizontal Bar Chart📈 Line Chart⭕ Scatter Plot🌡️ Heatmap📦 Box Plot📋 Histogram🥧 Pie / Donut Chart
DATA ANALYTICS SERIES:← Ch 9: Power BICh 10: Visualisation ←Ch 11: EDA →

The One Question That Determines Your Chart

Before choosing a chart, answer this question: what does the reader need to understand from this visual? Every chart type is built to answer a different kind of question. Match the chart to the question — not to what looks impressive or what you know how to make.

How do categories compare?
→ Bar Chart
How does something change over time?
→ Line Chart
Is there a relationship between two variables?
→ Scatter Plot
What is the part of a whole?
→ Pie / Donut (3–5 slices)
What is the pattern across two dimensions?
→ Heatmap
What is the distribution of values?
→ Histogram / Box Plot

5 Design Rules Every Analyst Must Follow

1. Start bar chart axes at zero

Truncating the Y axis (starting at ₹40L instead of ₹0) makes a small difference look enormous. Always start bar and column chart Y axes at zero. Line charts can use a non-zero baseline when showing small fluctuations in a trend, but be explicit about it.

2. Use colour intentionally

Colour should encode meaning — not decorate. One colour for all bars is correct when they represent the same thing. Use a second colour only when it encodes a difference (this bar vs that bar, this year vs last year). Rainbow palettes for categories look impressive but are harder to read than 2 well-chosen colours.

3. Write a specific title

"Revenue Chart" is useless. "Monthly Revenue — August 2026 (Delivered Orders, Noida Region)" answers the reader's first question before they even look at the chart. A good title states: what metric, what time period, what filters apply.

4. Remove chart junk

Gridlines, borders, 3D effects, shadows, background images, and decorative elements add visual noise without adding information. Every element that does not help the reader understand the data should be removed. Simpler charts are more honest and more persuasive.

5. Never use 3D charts

3D perspective distorts the visual representation of values. The slice at the front of a 3D pie appears larger than it is. The bars at the back of a 3D column chart appear shorter. 3D serves no analytical purpose and actively misleads — avoid it entirely in professional work.

📊

Bar Chart (Vertical / Column)

Use for: Comparing discrete categories
TYPICAL BUSINESS QUESTION: “Which city has the highest revenue?”
✓ USE WHEN
  • Comparing quantities across separate categories (cities, products, departments)
  • Ranking items — sort bars from highest to lowest
  • Up to 8–10 categories comfortably
✗ AVOID WHEN
  • Time series with many data points — use a line chart
  • More than 10 categories — use a horizontal bar chart or filter to top N
INDIAN BUSINESS EXAMPLE

Revenue by city (Mumbai ₹4.2Cr, Delhi ₹3.8Cr, Noida ₹2.1Cr…) — sorted descending so ranking is immediately visible.

RULES FOR THIS CHART
  • Always start Y axis at zero — truncated axes exaggerate differences
  • Sort bars by value unless the order itself is meaningful (Jan→Feb→Mar)
  • Use one colour for all bars unless colour encodes a different variable
  • Add data labels on top of bars for precise reading when exact values matter
📉

Horizontal Bar Chart

Use for: Comparing many categories or long labels
TYPICAL BUSINESS QUESTION: “Which product category has the highest return rate?”
✓ USE WHEN
  • Category names are long and would overlap on a vertical chart
  • More than 8–10 categories
  • Showing a ranked list (Top 20 products, Bottom 10 cities)
✗ AVOID WHEN
  • Time series — time should always flow left to right on a line chart
  • Fewer than 5 categories — vertical bars work fine
INDIAN BUSINESS EXAMPLE

Top 15 product sub-categories by return rate — long names like "Large Appliances", "Mobile Accessories" fit cleanly on horizontal bars.

RULES FOR THIS CHART
  • Sort by value descending (highest at top) unless showing a ranking that has a meaningful direction
  • For return rates or negative metrics, sorting ascending (worst at top) draws attention to the problem
📈

Line Chart

Use for: Trends over time
TYPICAL BUSINESS QUESTION: “How has monthly revenue changed over the past 12 months?”
✓ USE WHEN
  • Time is on the X axis (months, weeks, days, years)
  • Showing a continuous trend — revenue growth, user growth, stock levels
  • Comparing trends across 2–3 groups (this year vs last year)
✗ AVOID WHEN
  • Discrete categories — connecting "Noida to Mumbai" with a line implies a trend that does not exist
  • Too many lines (5+) — the chart becomes unreadable
INDIAN BUSINESS EXAMPLE

Monthly GMV trend Jan–Aug 2026 with a dotted reference line for target GMV — viewer immediately sees which months hit target and which did not.

RULES FOR THIS CHART
  • For 12 months of data, add data point markers (dots) so individual months are easy to identify
  • For daily data with high frequency, smooth the line only if the underlying trend is smooth — jagged daily data should stay jagged
  • Label the end of each line directly rather than using a legend when comparing 2–3 series
⭕

Scatter Plot

Use for: Relationship between two numeric variables
TYPICAL BUSINESS QUESTION: “Is there a relationship between delivery time and return rate?”
✓ USE WHEN
  • Showing correlation or relationship between two numeric variables
  • Identifying clusters or outlier data points
  • Each data point represents one entity (one city, one product, one customer)
✗ AVOID WHEN
  • Categorical data — bar chart is clearer
  • More than 1,000 points without additional encoding — overplotting makes it unreadable (use a heatmap instead)
INDIAN BUSINESS EXAMPLE

Cities plotted as points: X axis = average delivery time (days), Y axis = return rate (%). Cities in the top-right quadrant (slow delivery, high returns) are the problem areas to investigate.

RULES FOR THIS CHART
  • Add a trend line (regression line) to make the correlation direction explicit
  • Label outlier points directly on the chart
  • Size the dots by a third variable (e.g. revenue) to add a dimension without adding a chart
🌡️

Heatmap

Use for: Pattern across two categorical dimensions
TYPICAL BUSINESS QUESTION: “Which city × category combination drives the most revenue?”
✓ USE WHEN
  • Two categorical dimensions (city × category, month × product, hour × day)
  • The colour intensity communicates the magnitude quickly
  • Identifying patterns, hot spots, or gaps at a glance
✗ AVOID WHEN
  • Precise values are more important than the pattern — use a table or matrix
  • Only 2–3 categories — a bar chart is simpler
INDIAN BUSINESS EXAMPLE

Revenue heatmap: rows = 8 cities, columns = 6 categories. Dark orange cells = high revenue combinations. A pale cell in Noida × Large Appliances immediately signals an opportunity or a supply gap.

RULES FOR THIS CHART
  • Use a sequential colour scale (light → dark) for single-metric heatmaps — diverging scales (red → white → green) are for metrics that can be positive or negative (MoM growth)
  • Always include a colour legend with clear min/max labels
  • Add text values inside cells when the pattern is important but exact numbers also matter
📦

Box Plot (Box and Whisker)

Use for: Distribution and outliers in a numeric variable
TYPICAL BUSINESS QUESTION: “What is the spread of order values across product categories?”
✓ USE WHEN
  • Showing the distribution of a numeric variable — not just the mean
  • Comparing distributions across groups
  • Identifying outliers explicitly
✗ AVOID WHEN
  • Non-technical audience unfamiliar with the chart — use a histogram or just show median + IQR in a table
  • Fewer than 20–30 data points — the box has no statistical meaning
INDIAN BUSINESS EXAMPLE

Order value box plots for Electronics, Clothing, Home Decor, FMCG — the Electronics box spans ₹500–₹15,000 with many high-value outliers (phones), while FMCG has a tight box around ₹200–₹800 with few outliers.

RULES FOR THIS CHART
  • The box spans the IQR (Q1 to Q3). The middle line is the median. Whiskers extend to 1.5×IQR. Points beyond the whiskers are outliers.
  • Pair with a strip plot (dots overlaid) when the sample size is small enough to show all points
  • Always explain the chart to non-technical audiences — most people have never seen one
📋

Histogram

Use for: Distribution of a single numeric variable
TYPICAL BUSINESS QUESTION: “How are order values distributed — are most orders small or large?”
✓ USE WHEN
  • Understanding the shape of a distribution — normal, skewed, bimodal
  • Identifying where most values fall (the mode)
  • Checking for data quality issues (unexpected spikes or gaps)
✗ AVOID WHEN
  • Comparing across categories — use side-by-side box plots instead
  • Showing time trends — use a line chart
INDIAN BUSINESS EXAMPLE

Histogram of order values for an Indian D2C brand: most orders cluster around ₹400–₹900 (the mode), with a long right tail of high-value orders above ₹3,000 — classic right-skewed distribution common in e-commerce.

RULES FOR THIS CHART
  • Choose bin width carefully — too few bins hides the shape, too many bins creates noise. For order values, ₹100 bins typically work.
  • Add a vertical line at the median to show where the centre is — the mean is often misleading in right-skewed distributions
  • Histograms look like bar charts but are different — bars in a histogram are adjacent (no gap) because the variable is continuous
🥧

Pie / Donut Chart

Use for: Parts of a whole — for 3–5 categories only
TYPICAL BUSINESS QUESTION: “What share of total revenue comes from each major channel?”
✓ USE WHEN
  • 3–5 categories that sum to a meaningful whole (100%)
  • The question is explicitly about proportions
  • One slice is dramatically larger than the others — the visual impact is clear
✗ AVOID WHEN
  • More than 5 slices — use a bar chart
  • Values are similar in size — the slices look identical and the chart adds no clarity
  • Precise comparisons are needed — use a bar chart where length is easier to judge than angle
INDIAN BUSINESS EXAMPLE

Revenue by channel: Direct 58%, Marketplace 28%, Wholesale 14% — three slices with meaningfully different sizes that clearly show Direct dominance.

RULES FOR THIS CHART
  • Never use 3D pie charts — perspective distorts slice sizes
  • Start the largest slice at 12 o'clock (top) for conventional reading
  • When in doubt between a pie and a bar — use the bar

Quick Reference — Chart Selection at a Glance

Chart TypeBest ForAvoid WhenMax Categories
Bar (vertical)Comparing ≤10 categoriesTime on X axis10
Bar (horizontal)Long labels, many categories, rankingTime trends20+
LineTime trends, continuous dataDiscrete categories3–4 lines
ScatterRelationship between 2 numeric variablesCategorical dataN/A (points)
Heatmap2-dimensional patterns (matrix view)When exact values matter mostUnlimited
Pie / DonutParts of a whole, very few slices>5 slices, similar-sized values5
HistogramDistribution of one numeric variableComparisons across groupsN/A (bins)
Box PlotDistribution + outliers across groupsSmall samples (<20), non-technical audience6–8 groups
Continue the Series
← Ch 9: Power BICh 11: EDA Complete Guide →

Frequently Asked Questions

What is the most important rule in data visualisation?

The most important rule in data visualisation is: the chart should answer one specific question clearly, without the reader having to study it. Every design decision — chart type, colour, axis range, title — should serve that single purpose. A chart that requires a legend, a footnote, and a verbal explanation has failed. A chart that a busy manager can read in 5 seconds has succeeded. The second most important rule: do not distort the data. Truncated Y axes, 3D charts, and dual axes with different scales all make data look different from what it actually is — either accidentally or deliberately. Good analysts never mislead with charts, even unintentionally.

When should you use a bar chart vs a line chart?

Use a bar chart (vertical columns or horizontal bars) when comparing discrete categories — revenue by city, orders by product category, headcount by department. The categories are separate things with no inherent connection between them. Use a line chart when showing change over time — monthly revenue, daily orders, weekly user growth. The key distinction is continuity: a line implies that values between the plotted points are meaningful (the revenue was somewhere between January and February values throughout the month). If your X axis is time, use a line. If your X axis is a list of separate things (cities, products, salespeople), use a bar chart. A common mistake is using a line chart for category comparisons — lines between cities (Noida → Mumbai → Delhi) imply a trend that does not exist.

When is it acceptable to use a pie chart?

Use a pie chart only when: (1) you have 3–5 categories maximum — more than 5 slices becomes unreadable; (2) the proportions are meaningfully different — if all slices are 20%, a pie is pointless; (3) the question is specifically about parts of a whole — "what percentage of our revenue comes from each category"; (4) you do not need precise comparisons — pie charts make it hard to compare slice sizes accurately. In most cases, a horizontal bar chart is clearer than a pie chart. Pie charts are appropriate for the occasional "one big vs many small" situation — "Direct channel drives 65% of revenue vs all other channels at 35% combined." For anything more nuanced, use a bar chart.

What are the most common data visualisation mistakes made by analysts?

The most common data visualisation mistakes made by Indian data analysts in professional settings: (1) Truncated Y axis — starting the Y axis at something other than zero makes small differences look large. Always start bar chart Y axes at zero. (2) Too many colours — using a different colour for every bar or category creates visual noise. Use one colour for all bars unless colour encodes something meaningful. (3) No title or vague title — "Revenue Chart" is useless. "Revenue by City — August 2026 (Delivered Orders Only)" tells the reader what they need. (4) Using 3D charts — 3D makes charts harder to read and distorts proportions. Never use 3D in analytical work. (5) Too many visuals on one page — a dashboard with 20 charts is as useless as a report with 200 pages of data. Fewer, clearer visuals are always better. (6) Pie charts with 8 slices — unreadable. Use a sorted bar chart instead.

EVIKA ACADEMY · NOIDA SECTOR 51

Build Dashboards That Communicate Clearly

Our curriculum includes visualisation projects in Excel, Power BI, and Python — where you practice choosing the right chart for real Indian business questions and presenting findings to a mock management audience.

Book Free Demo Class →
🎓 Free Demo Class — Online & Offline · Noida Sector 51