Hectar’s Audrey Bourolleau and Francis Nappez presented their findings about greenhouse gas emissions in the industrial production of bread baguettes at the 2024 Lean Summit in France. They see a major impact in (1) farming and (2) the production of fertilizer and plant protection products. Together, these categories account for 58% of total emissions but barely 6% of the costs. This suggests that improvements in these two areas could cut emissions in half with a minimal impact on bread prices.
This is about the visualization of this kind of information with bump charts/slopegraphs. Edward Tufte prefers slopegraph but bump chart is more common.
Bourolleau and Nappez used the following chart to make their point:
It buries the lede. First, the picture in the middle misled me into believing that the numbers pertained to different sections of the baguette. It is decoration only, separating numbers you want to connect and contrast.
The names of the different stages of the baguette supply chain are in small print below the Cost line instead of being highly visible at the top of the column. Finally, the detailed text explanations are unnecessary for operations outside the presenters’ mission to improve agriculture at Hectar.
Alternative Visualization
I would have used a less colorful chart, more focused on the key point:
Bump Charts
The above is a bump chart, also known as a slopegraph, a common method for representing categories that rank differently based on different criteria. This chart uses color to focus attention on two categories, and the explanations are included only for these two.
These charts are commonly used, but do not receive much attention in the literature on visualization, perhaps because they are self-explanatory. In Product-Quantity (PQ) analysis, for example, you rank products by volume, but the rankings will vary with whether you are measuring quantities in sales, numbers of units, weight, or bulk, and you may use bump charts to compare any two.
Bump Chart of Nissan Production Sequence
In the Japanese book on 25 keywords of the Nissan Production Way, the following chart is used to explain their approach to sequencing performance in terms of numbers of cars moved up or down:
Tour de France Runner Rankings
It is also used to show how rankings change over different time periods. In the discussion of Rank-and-Yank, this diagram shows how the worst performer in the winning team of the 2011 Tour de France, Marcus Burghardt, performed much better in 2012:
Multiple Rankings
It can also be used for more than two rankings, but it can quickly lose clarity as the number of items and rankings rises. The following is an example generated with Tableau by Adrien Sourdille about sales of product families by period:
Bump Chart Design Guidelines
To keep the chart informative, you can enhance it in several ways:
Put the category labels next to the first and last columns rather than in a legend box to the side.
Label the x-axis.
Use color to highlight one or two of the categories, leaving all others grayed.
G20 Carbon Dioxide Data
This example, from an article on LinkedIn Pulse, follows recommendations 1 and 2 above, but the color scheme still makes the chart difficult to read, with lines of the same color crossing. To emphasize that the US stayed on top while the UK dropped from 2nd to 10th place and Saudi Arabia rose from 14th to 2nd between 1980 and 2010, for example, you could gray out all the other countries.
JD Power Initial Quality Survey of Car Brands in the US Market
This is an example based on the JD Power Initial Quality Surveys of the US car market over 20 years. It only includes brands that have been reviewed throughout this period, and therefore excludes the current Genesis brand at the top and electric car makers Tesla, Rivian, and Polestar at the bottom. Using colors, we can focus attention on the drop in rank by Lexus and Toyota:
The above chart showed ranks only. We can also locate the points based on the value of the metric used for ranking, the number of problems reported per 100 vehicles by owners within the first 90 of ownership of a new car:
It shows a market-wide decline in absolute values, along with a drop in rankings for Lexus and Toyota. Whether initial car quality is worse or customers are more demanding, this metric has been moving in the wrong direction.
In R, bumpchart is a function within the plotrix package, that works with a sequence of rankings, provided there are no missing values. For the car brand rankings by initial quality, I couldn’t coax it to show brands that vanished or appeared between 2004 and 2024. The ggbump package is an alternative.
Conclusion
Bump charts, or slopegraphs, are another useful tool in visualization panoply. The above examples show how to use them effectively, as well as pitfalls to avoid.
References
日産生産方式キーワード25 (2005). Nikkan Kogyo. (25 keywords of the Nissan Production Way)
Jul 21 2024
Rankings and Bump Charts
Hectar’s Audrey Bourolleau and Francis Nappez presented their findings about greenhouse gas emissions in the industrial production of bread baguettes at the 2024 Lean Summit in France. They see a major impact in (1) farming and (2) the production of fertilizer and plant protection products. Together, these categories account for 58% of total emissions but barely 6% of the costs. This suggests that improvements in these two areas could cut emissions in half with a minimal impact on bread prices.
This is about the visualization of this kind of information with bump charts/slopegraphs. Edward Tufte prefers slopegraph but bump chart is more common.
Contents
Visualization in the Presentation
Bourolleau and Nappez used the following chart to make their point:
It buries the lede. First, the picture in the middle misled me into believing that the numbers pertained to different sections of the baguette. It is decoration only, separating numbers you want to connect and contrast.
The names of the different stages of the baguette supply chain are in small print below the Cost line instead of being highly visible at the top of the column. Finally, the detailed text explanations are unnecessary for operations outside the presenters’ mission to improve agriculture at Hectar.
Alternative Visualization
I would have used a less colorful chart, more focused on the key point:
Bump Charts
The above is a bump chart, also known as a slopegraph, a common method for representing categories that rank differently based on different criteria. This chart uses color to focus attention on two categories, and the explanations are included only for these two.
These charts are commonly used, but do not receive much attention in the literature on visualization, perhaps because they are self-explanatory. In Product-Quantity (PQ) analysis, for example, you rank products by volume, but the rankings will vary with whether you are measuring quantities in sales, numbers of units, weight, or bulk, and you may use bump charts to compare any two.
Bump Chart of Nissan Production Sequence
In the Japanese book on 25 keywords of the Nissan Production Way, the following chart is used to explain their approach to sequencing performance in terms of numbers of cars moved up or down:
Tour de France Runner Rankings
It is also used to show how rankings change over different time periods. In the discussion of Rank-and-Yank, this diagram shows how the worst performer in the winning team of the 2011 Tour de France, Marcus Burghardt, performed much better in 2012:
Multiple Rankings
It can also be used for more than two rankings, but it can quickly lose clarity as the number of items and rankings rises. The following is an example generated with Tableau by Adrien Sourdille about sales of product families by period:
Bump Chart Design Guidelines
To keep the chart informative, you can enhance it in several ways:
G20 Carbon Dioxide Data
This example, from an article on LinkedIn Pulse, follows recommendations 1 and 2 above, but the color scheme still makes the chart difficult to read, with lines of the same color crossing. To emphasize that the US stayed on top while the UK dropped from 2nd to 10th place and Saudi Arabia rose from 14th to 2nd between 1980 and 2010, for example, you could gray out all the other countries.
JD Power Initial Quality Survey of Car Brands in the US Market
This is an example based on the JD Power Initial Quality Surveys of the US car market over 20 years. It only includes brands that have been reviewed throughout this period, and therefore excludes the current Genesis brand at the top and electric car makers Tesla, Rivian, and Polestar at the bottom. Using colors, we can focus attention on the drop in rank by Lexus and Toyota:
The above chart showed ranks only. We can also locate the points based on the value of the metric used for ranking, the number of problems reported per 100 vehicles by owners within the first 90 of ownership of a new car:
It shows a market-wide decline in absolute values, along with a drop in rankings for Lexus and Toyota. Whether initial car quality is worse or customers are more demanding, this metric has been moving in the wrong direction.
Generating Bump Charts
Generating a bump chart in Excel is manual work. Darren Gosbell has a method to build a bump chart in Power BI using Deneb.
In R, bumpchart is a function within the plotrix package, that works with a sequence of rankings, provided there are no missing values. For the car brand rankings by initial quality, I couldn’t coax it to show brands that vanished or appeared between 2004 and 2024. The ggbump package is an alternative.
Conclusion
Bump charts, or slopegraphs, are another useful tool in visualization panoply. The above examples show how to use them effectively, as well as pitfalls to avoid.
References
#bumpchart, #slopegraph, #visualization
Share this:
Like this:
Related
By Michel Baudin • Data science • 0 • Tags: Bump chart, Slopegraph, Visualization