Hands-on Tutorials
This is a carte figurative. It was created by Charles Joseph Minard to show Napolean’s Russian campaign of 1812–13.

Modern information scientists say that it is the best statistical graph ever produced in the history of mankind. Minard’s graph tells a rich, coherent story from multivariate Data and depicts the size of the army, location in a 2-D plane, direction of movement, temperature on various days during the retreat. It is one of the most succinct displays of data using a visual.
Can we create our own Carte Figurative?
Maybe yes!
The modern tech and scientific publication has 25% visual info, the decks we create for our day to day business activities have around 60% visual content (graphs, diagrams etc); It makes a solid case for movement towards graphical excellence.
Why focus on Visualisation?
Data visualisation is a major part of Data Science and Machine learning as one has to explore the data before putting it in models and later communicate the results to relevant stakeholders.
Visual Display of Quantitative Information is a timeless classic by Edward Tufte (pronounced as Tuftee) and one of the best non-fiction books of the 20th century. It talks about the history of visual graphics, precepts of graphical excellence, and fundamental rules to follow while designing data-based graphics.
As a data scientist, I have benefitted greatly from the knowledge that’s present in 200-odd pages.
Key takeaways and towards better charts
- Remove clutter – Remove everything that you don’t need. Less is more.
- No pie charts – They don’t do much as the human eye can’t decipher the size of the slices of a circle esp. when they are somewhat equal.
- Don’t need visual every time – If you don’t have enough data then use only a large number or a table to display the information.
- Keep the audience in mind – The audience shouldn’t spend a lot of time parsing the visual. The data should stand out from the visual itself. Use the basic and primitive visual cues such as colour, shape, size to direct the attention.
- Don’t distort the data – The concept that torture the data enough to tell your story is not a good principle to follow. Let the data speak for itself.
How to design better charts?
That’s where graphical excellence comes in.
Graphical excellence gives the viewer the greatest number of ideas in the shortest time in the smallest space.
You can achieve this in your charts via three simple mechanisms:
- No chart junk – Remove weird fonts, textures, shadings or anything that can distract the audience from the central idea of the visual and data.
- No useless information – Large share of your graph should be used in displaying the data.
In simple words, show only what’s necessary.
- Substance over design – Your chart should induce the user to think about the data and the substance that it is offering rather than getting lost in the methodology behind producing the graphic/visual. So, appeal to the natural way of how humans understand and direct their attention – from large to small, from dark to light, from left to right, from up to down.
The other important point for the creation of the visual is the intent – why the visual was created in the first place? The content of the chart should be designed in a manner that should drive the decisions by communicating conclusions.
The driving force behind your chart shouldn’t be the use of certain tools, software packages, or libraries but to communicate conclusions you made from the data.
Charts in action
Let’s see how the principles can be applied in our work.
I created some dummy data for films I watched and liked between 2011 to 2016. Let’s create a year-on-year comparison chart.
I came up with this in the first iteration.

Let’s see if we can improve on this.
It seems there is a lot of chart junk. The first one is the table below the chart, the axis labels. There is a dense background grid that doesn’t serve any purpose and is thus disposable.
A little change here and there and we have come to a much better visual than before.

It is much cleaner, crisper, and is able to depict what it wants to without presenting any unnecessary information.
One more example
A fictitious store owner wants to compare the sale of various commodities first half of the year and he created the following viz.

There isn’t anything fundamentally wrong with this visual but there isn’t any conclusion that is standing out. What’s the pattern here? What should the store owner conclude?
Could this be a better way to visualise?

By breaking the chart 4 ways, a pattern is observed and that is the key takeaway.
Ok, one more!
A new product has been launched by a fictitious company and the marketing team has surveyed existing customers.

It tells us that most of the users aren’t interested but that’s that and the rest half of the pie isn’t too informative.
This can and should be remedied by sorting the data.

This seems to be a much better and an informative visual as the reader can understand very quickly that who is least interested and who is most interested in the product.
Make information beautiful
The goal of the data based graphs isn’t to create visuals for the sake of creation but for communicating conclusions so as to drive decisions.

A visual can help you greatly if you take a step back, think what’s the purpose of the graphic, keep things simple while designing, and have a story to tell.