Show Me the Numbers - Book review
1 June 2021
Dr. Moria Levy
"Show Me the Numbers: Designing Tables and Graphs to Enlighten" is the definitive guide for effectively presenting tables and graphs. Authored by Stephen Few in 2011, the book meticulously explores every aspect of the domain of tables and graphs, delving into even the minutest details, such as the incorporation and precise placement of tick marks.
According to Few, tables and graphs have four primary goals: analysis and learning, communication, tracking, and forecasting. This book focuses explicitly on the realm of communication. For those with interests in other aspects, Few addresses, at least to some extent, those needs in his other works.
For individuals who may be undecided, it becomes evident that there is a wealth of knowledge to acquire. Often, we unknowingly engage in nonsensical practices—numbers, by themselves, fail to articulate a clear message unless we guide them correctly.
The book covers the following subjects:
Tables vs. graphs
Types of tables
Types of graphs
Shared recommendations for visual clarity and cognitive understanding
Crafting a narrative based on the data
The summary provided below distills the core practical recommendations. It deliberately excludes theory, explanations (such as defining a line, point, or change), rationales, exceptions, and examples. It gets straight to the essentials. I highly recommend reading the summary first and then delving into the book. It is highly recommended!
Tables vs. graphs
Tables are structured in a column-row format to organize and present information, displaying data in rows and columns interpreted through text. Graphs, on the other hand, serve as a visual means of expressing quantitative data. Data is portrayed within an area delineated by one or more axes, with these axes providing a numerical or categorical scale for marking and situating values relevant to the visual elements.
Criteria for preferring the use of tables:
Non-quantitative data presentation: When the data lacks a quantitative nature, it is unsuitable for graph representation (e.g., a seminar schedule).
Specific data view and review: When examining and revisiting particular data points, it is necessary.
Comparing specific data from a large dataset, Especially when the objective is focused on particular comparisons rather than an exhaustive analysis.
Accuracy of values: When precise values are essential.
Multiple units of measurement: When the quantitative information encompasses more than one unit of measurement.
Requirement for detailed and summary information: This becomes particularly relevant in the subsequent story chapter on Types of Summary Information.
Benefits of Tables:
Ease of understanding: Human beings have utilized tables for over two thousand years, making them familiar and straightforward.
No brain translation required: The table information is inherently precise, requiring no additional mental processing.
Criteria for opting for a graph:
Comprehensive comparative presentation: The aim is to provide an all-encompassing close view of a dataset.
Illustrating relationships within and between value groups: Particularly useful when emphasizing the form or structure of these relationships.
Message conveyed through visual appearance: The essence of the message lies in the visual attributes of the values, such as patterns, trends, or anomalies.
Benefits of Graphs:
Ease of understanding: Graphs, having been employed by humanity for two thousand years, remain a universally accessible means of communication.
Remember, in cases where the narrative is straightforward and entails presenting just one or two figures, it might be more effective to utilize text, construct a sentence, and bypass the use of graphs and tables.
Types of tables
There are several main types of tables:
One-way tables: Lists where item categories are displayed in only one direction, either columns or rows (e.g., a list of products with the number of sales and profits per product). While these tables can include multiple categories (e.g., district and product categories), the information is still organized in the same direction. In the case of multiple categories, generating a breakdown or completeness or creating a hierarchy (e.g., district and city) is possible, but this does not affect the direction; it influences groupings and breakdowns.
Bidirectional tables: Matrices in which each table value is attributed to a specific column and row, representing a category. The advantage of two-way tables is their space-saving nature. They can also accommodate hierarchies of categories.
Select the table type:
One Category - One-Way Tables
Two categories (without hierarchy) – Exercise discretion; Two-way tables save space.
One or more categories with hierarchy – Exercise discretion; One-way tables provide a more precise representation.
Types of graphs
Here are various typical needs and graphs appropriate for presenting them:
Point Scattering Chart: Inappropriate when using a bar graph due to a measurement scale that doesn't start with zero.
Vertical or Horizontal Bar Graph: Suitable for numerical comparisons.
Point Scattering Chart: Only when values aren't collected at consistent times.
Line Graph: Emphasizes an overall trend.
Bar Graph: Emphasizes individual values.
Box: Appropriate only when illustrating the distribution of the variable over time.
Point Scattering Chart: Suitable when a bar graph is inappropriate due to a non-starting zero scale.
Vertical or Horizontal Bar Graph: Suitable for rating comparisons.
Boxes: Appropriate only when ranking distribution groups.
Wired Graph: Used to illustrate changes in trends over time.
Vertical or Horizontal Bar Graph: Suitable for paying parts.
Point Scattering Chart: Appropriate when using a bar graph is unsuitable due to a non-starting zero scale.
Line Graph: Suitable for describing deviations in time series.
Vertical or Horizontal Bar Graph: Suitable, always vertical when combined with time series.
Strip Plot Diagram: Corresponds to single or multiple distributions.
Linear Graph – Polygon: Single distribution – emphasis on an overall trend. Multiple distributions – limit to a few lines.
Histogram: Suitable for a single distribution focusing on a few intervals.
Plot Box: Appropriate for multiple distributions only.
Scatter Plot Diagram: Suitable for illustrating correlation.
Vertical or Horizontal Bar Graph: Suitable in a tabular lens format.
Point Chart: Suitable, using different point sizes to indicate values.
Line Graph: Used to mark routes.
*Note: In graphs for this purpose, the X-axis represents categories, and the Y-axis represents numerical values.
Boxes are a lesser-known graph type akin to ungrounded bar graphs. They display distribution groups, including bottom range, top (bottom and top of the box), and mean (marked with a stripe).
It's worth noting that many types of charts are intentionally omitted due to their potential for misleading information. Notably absent is the pie graph, which was excluded due to its tendency to mislead. Additionally, 3D graphs are not used as they are challenging to read accurately and don't enhance understanding.
Despite incorporating visual elements, tables primarily engage with the textual system in our brains. Consequently, we read them sequentially, from right to left and top to bottom (in English), somewhat in reverse. Ensure, to the extent possible, that crucial messages, if any, are placed at the outset of the table.
a) Prefer white space where feasible, followed by subtle colors and fine stripes if necessary. Avoid using a grid.
a) Organize columns or rows into categories that fit within page/screen widths in separate columns. Time should always be horizontal, series in separate columns, and cascading in portrait orientation.
b) Maintain sufficient space between groupings for clear separation. Ensure a uniform table structure between groups. Consider placing each group on a new page when groups require separate reading.
c) In columns, list the names of categories referring to the quantity to the left of them. Register categories with left-to-right hierarchies in a way that reflects the hierarchy. Calculated data should immediately appear to the right of the data on which it is based.
d) Sorting: When order is essential, sort accordingly.
a) It's best to avoid non-horizontal directions.
b) Always align numbers to the left, maintaining the same number of decimal digits as the table. Align dates however desired, ensuring uniformity in different values (including the number of characters). Center columns with the same number of characters are more minor than the heading (e.g., gender M/F).
c) Format numbers using a comma to indicate thousands (every three digits). Consider chopping digits (by triples, thousands, millions, etc.) as much as possible and appropriate. When describing negative numbers, use parentheses, aligning the internal content with positive ones.
d) Dates: Days in 2 digits, months in 2 digits or three characters, and years in 2 or 4 digits - maintaining uniformity.
e) Choose a readable font. Use a single font.
f) Accents and color – utilize bold, italic, or font color change (italics are discouraged).
Summary of Values:
a) Position summary values at a visual distance from detailed data to draw attention.
a) Always repeat column headings and row headings (e.g., groupings) at the beginning of each page.
Graphs are Visual: Graphs primarily interact with the brain's visual systems, which lack a fixed reading order. Attention is drawn to what stands out most in the diagram, necessitating considering elements like color, cropping points, and size. These aspects play a central role in conveying the message effectively.
Representation of Variable Size: It is advantageous to utilize variable length or position in two-dimensional space to represent a variable size, with the latter being preferable. Other methods of representing varying sizes can be misleading.
Perceptual Limits: We can adequately distinguish up to 8 shades, four directions, and about four sizes. It is advised not to exceed these limits.
Recommendations Related to Components of the Graph:
Points: Increase the size or choose more visually distinct shapes when there isn't enough distinction between groups of points. Overlapping points can be corrected by adjusting the graph size or decreasing each point. Additionally, reducing fill colors to dots can help when points overlap.
Bars, Columns, Boxes: Utilize horizontal columns with limited space for text labels. Avoid horizontal columns for time series. Maintain a similar width for white separators between bars, and avoid overlapping bars. Use balanced colors with apparent variations between them. Accentuate a bar using color more prominently than others. Columns always start from a scale of 0.
Strokes: Separate different shades of lines whenever possible. Lines should include points only if emphasizing the same point on other lines.
Combinations: Combine boxes and lines for describing distributions over time. Bars and points are suitable for indicating comparisons when there is no congestion.
Trend Lines: Prefer using variable average scores rather than an overall average line, which can be misleading. General linear average lines are appropriate only in scatter plot charts.
Reference Lines: Suitable for specifying significant and bounded thresholds, particularly in measuring the norm.
Annotations: Text accompaniment is recommended to illuminate and explain specific values, supporting the overall message.
Log Scales: Appropriate for reducing visual differences between quantities with significant differences in value groups. Measurement contracts can be helpful in these graphs.
Ticks: Inconspicuous about data. It can be combined with numeric values but not with categories—balance quantity/density to avoid overloading the scale. Ensure the use of round numbers in ticks.
Gridlines: Subtle, fine in color, and used only when reading or comparing data.
Legend: Appropriate when labels represent categories and cannot appear on the axis itself. Located close to the data it represents. Inconspicuous about the data. Boundaries can be helpful when separating from other information.
Hinges: Avoid playing with aspect ratios.
Data Area: Keep backgrounds clean and light.
Combining Several Graphs: Maintain design consistency between graphs. Exceptions include titles and legends, which can be avoided. Organize the order of graphs logically for easy focus and substantive comparisons. Use rulers to separate graphs only if necessary to distinguish them clearly (e.g., column or matrix format).
Shared recommendations for visual clarity and cognitive understanding
There is a strong correlation between what is visually accessible for us to perceive and what is easy for us to comprehend.
Principles of Operation (Article 3-5 - Gestalt Principles):
Visual characteristics such as size, length, color, etc., are all influenced by context. It is essential to be aware and apply common sense by using combinations that are clear and not misleading.
Colors that work well together and are easy to identify include gray, blue, orange, green, pink, brown, purple, yellow, and red. When combining different colors, it is recommended that their intensities be similar. The size of objects also influences the ability to identify them—the smaller they are, the more challenging it is for us to distinguish between them.
Elements close to each other are perceived as associated with a joint group.
Elements similar in color, shape, or direction are perceived as associated with a joint group.
Elements bounded by a background or frame are perceived as associated with a joint group. It is sufficient to mark x, y-axes, or a partial frame in a table to convey the association of content, reducing unnecessary graphic clutter.
Highlight important information to communicate powerfully and clearly without interruptions—reducing ink usage for unimportant data. Add emphasis to essential data in your message. Characteristics that can be affected include line width, direction, size, differentiation, hue, and color intensity.
Organize information thoughtfully to enhance understanding and use.
Group information into standard segments using categories.
Prioritize and rank importance using the highlighted elements as indicated above.
Set an order for reading—internally in tables by arrangement and in graphs by organizing a series of graphs.
Combine the use of graphs and text tables for the most effective message. Texts can serve as labels, introductions, explanations, emphasis, and repetition to reinforce a message, as an order, as a recommender, and as a tool to raise questions. Text can answer questions like what, when, who, and where.
Crafting a narrative based on the data
As mentioned earlier, this book's primary focus is effectively communicating data through tables and graphs. The default perception is that numbers are often considered "boring," and telling a successful story and conveying a message through exhibits requires a dedicated investment. In addition to the specific recommendations provided earlier, the following are guiding principles for creating a compelling story:
Simplicity and Relevance: Remove anything unnecessary. A good graph, table, and content should precisely include what is essential and nothing more.
a) A data-based story describes relationships in quantitative information, such as the correlation between a specific issue and the number of days it takes to resolve. Relationships can be simple to complex, including ranking, ratios, or correlations.
b)Data can be raw or summarized (e.g., sum, average, median, range, variance, etc.).
Critical Principles for a Successful Story: Keep it simple, flowing, inclusive, accurate, contextual, familiar, concrete, personal, emotional, in order, and motivating for action.
Comparison in Tables and Graphs: Utilize comparisons to convey the message effectively, comparing against defined goals, forecasts, other objects (companies, products, etc.), norms, and other periods.
Text and Visual Integration: A compelling story often combines text and images with graphs and tables. The latter should support the primary message you aim to convey.