You are viewing a public preview
Sign in to save your progress, do assignments, join discussions and get feedback.
Concept, purpose and role of data visualization; the relationship between business questions, data, information and decisions; benefits and limitations of visual representations. The slides compare good and bad visualizations and walk through the common pitfalls that mislead readers.
Learning outcomes
The lecture slides for this lesson are below.
By the end of this lesson you can:
Key idea: A business chart exists to help someone decide. It must show the data faithfully, make the important pattern easy to see, and never distort what the data say.
Data visualization is the representation of data with visual elements such as position, length, colour and size, so that people can see patterns, comparisons and changes. The lecture gives two reasons to visualize:
The lecture shows this with a small table of orders by day of the week. Even with only seven numbers, you must read and compare every value to find the busiest and quietest days. In a bar chart the answer appears at a glance, because the eye compares bar lengths almost without effort.
In a business, visualization sits in a chain: a business question (for example, "where are we losing profit?") tells you which data to collect; analysis and charts turn the data into information; and information supports a decision. A chart that does not connect to a question rarely changes a decision.
The lecture lists four requirements:
The slides then compare good and bad examples. A good chart has a clear purpose, an honest scale, readable labels and little clutter. A bad chart uses decoration, inconsistent scales or too many colours, so the reader must work hard and may reach the wrong conclusion.
| Pitfall | What goes wrong | How to avoid it |
|---|---|---|
| Truncated y-axis / scale distortion | The axis starts far above zero, so small differences look huge | Start bar and column axes at zero; if a line chart zooms in, label the axis clearly |
| Data not adding up | Percentages in a part-to-whole chart total more or less than 100% | Check totals; use part-to-whole charts only when parts form one whole |
| Omitted data | Periods or categories are left out to make a trend look better | Show the full range, or state clearly what is excluded and why |
| Arbitrary dual y-axes | Two measures on two scales are made to look related by choosing the scales | Use dual axes only when there is a real reason, label both axes, or use two charts |
| Manipulated y-axis | Axis is inverted, stretched or uneven, so the shape misleads | Use a standard, evenly spaced axis with zero clearly marked |
| 3D distortion | Perspective makes front slices or bars look bigger | Use flat 2D charts |
Benefits: faster understanding, patterns and outliers become visible, easier communication across departments, and support for monitoring and decisions. Limitations: a chart shows less precision than a table; a poorly chosen chart can mislead; charts can hide how the data were collected or cleaned; and readers may see a cause where there is only a correlation. When people need exact values to look up, a table is still the right tool.
Use the lecture's table of orders by day of the week: Monday 964, Tuesday 653, Wednesday 482, Thursday 876, Friday 578, Saturday 932, Sunday 1108.
Flashcards · Overview of Data Visualization in Business
8 wordsRepresenting data with visual elements such as position, length, colour and size so that patterns and comparisons are easy to see.
An axis that starts well above zero, exaggerating small differences between bars.
Any choice of scale that makes visual size disagree with the real size of the values.
Leaving out periods or categories so that the remaining data tell a more convenient story.
Two measures on two independently scaled axes, which can create a false impression of correlation.
Perspective effects that make front slices or bars look larger than they are.
Showing data from a broad overview down to granular detail so readers can drill in.
A part-to-whole chart whose parts do not total 100% or the stated whole.