Say It with Charts : The Executive's Guide to Visual Communication (4th)

Full Title: Say It with Charts : The Executive's Guide to Visual Communication (4th)
Author: Gene Zelazny
ISBN/URL: 9780071369978
Reading Period: 2020.03
Source: Referred to in The McKinsey Way: Using the Techniques of the World's Top Strategic Consultants to Help You and Your Business

General Review

Read this book if you want to really communicate effectively with charts
Highly recommended
This book highlights many details about charting that you may not have paid particular attention to, but definitely have affected how you perceived a chart.

First, determine the message I want to get across
- Given the same set of data, I might want to bring across a different message. E.g., given market share data across multiple years, I might want to highlight the increase in market share of a certain company, or I might want to highlight that a certain company has small market share
- The message should be the title of the chart. E.g., instead of "Company Sales Trend", I should use something like "Company Sales Have Doubled"
Next, ascertain what kind of comparison does the message entails
- Component – When I'm interested in showing the size of each part as a percentage of the whole
- Item – When I'm trying to highlight whether something is more or less than other things; when ranking is important
- Time Series – When showing changes over time
- Frequency Distribution – When showing a count of items falling into various numerical ranges (think literally normal distribution, poisson distribution, etc.)
- Correlation – Relationship between variables
Finally, select the chart form
The table below shows the chart form(s) that generally goes well with each particular comparison type

Comparison Type Chart Form

Component Pie

Item Bar

Time Series Line or Column

Frequency Distribution Line or Column

Correlation Bar or Dot

Generally pie chart is appropriate because the angles effectively show how much each component contributes to the total (e.g., an angle of 90 degrees corresponds to 25%)
When showing change is components composition over time, consider using 100% bar or column charts, with lines connecting the components from one time period to another
Tricks
- Sometimes placing the component at the bottom (at around 4 o'clock to 8 o'clock position) has the effect of making the component look bigger (see example of Apple making 21.2% appear greater than 19.5% at https://www.wired.com/2008/02/macworlds-iphon/)

Generally a column bar chart is appropriate because:
- the labels can be written horizontally to the left of the bars (within the need to rotate the labels as in the case of a vertial bar chart)
- a single scale can be placed at the top
- this reduces the possibility of the item comparison being confused with time comparison (which is usually presented on a vertial bar chart where time flows horizontally)

Generally a column (fewer data points) or line chart (more data points) works best
Tricks
- For a column chart, to emphasize a particular column, try using a downward-pointing arrow just above that particular column
- For a line chart, instead of drawing multiple lines on one chart, split the data into multiple chart comparing just two items in each chart (e.g., instead of plotting trend across time of company A, B, C and D on one chart, trying plotting three charts – A v B, A v C, A v D, assuming A is the common basis of comparison)
Variations
- Instead of a simple line chart, consider Surface chart / subdivided surface chart (i.e., two or more line charts stacked vertically together)

Generally a column (fewer data points) or line chart (more data points) works best
Frequency distribution comparison can be thought of as:
- a time series comparison where the dimension of time is replaced with things like size of sales (e.g., plot of number of sales (y-axis) against size of sales in dollars (x-axis))
- an item comparison where the "items" are actually a binned groups of an otherwise single item (e.g., plot of number of employees (y-axis) against various age groups (x-axis))

Correlation comparison generally means to show whether a correlation exists or does not exist (e.g., whether size of discount correlates with number of sales)
Generally a scatter plot or paired bar chart is best
Variations to consider
- On a scatter plot, it is possible to include a third dimension as the size / colour of the dots

Where should the axes be (e.g., left or right, top or bottom)
Should the scale be absolute values, or index / percentage
How should the data points be ordered (where there is flexibility)
How to visually emphasize what is most relevant
- Consider shapes and bounding boxes like circles, triangles, stars, and arrows

Things I should put into my day-to-day toolbox include:
- When titling graphs, makes sure the title conveys a message
- Don't be constrained to the default graph forms; always think of possible variations

Whenever I see any data, try to think of ways to present it in a graphical manner

Browse through Section Two for example usages of various charts across all comparison types to get inspiration when considering on how to present data
- The example usages show how to emphasize particular data points