Book Review: The Visual Display of Quantitative Information – Edward Tufte
So here we have it; a book review of one of the two Qlikview recommended ‘must reads’ (the other being Stephen Few’s ‘Information Dashboard Design’ which I’ll cover at a later date), many of you will have already heard of the work of Edward Tufte so this review is perhaps aimed more at those new to Qlikview, Chart Design and Data Visualisation – although there is a crucial point for existing readers of this book at the end.
First things first; who’s this book aimed at and suitable for? In short it should be read by anyone who has even a passing interest in Data Visualisation, this isn’t simply due to the ‘high level’ nature of the book but also the fact that it’s written in a very accessible style – very much unlike the more traditional ‘how to’ text books a la Mr Few’s. This ready accessibility is what helped the book gain a place on Amazon’s list of the Top 100 Non-Fiction Books of The Twentieth Century and it’s also what makes it the ideal starting point for the relative novice.
But you don’t get on a list like that simply by being accessible; the accessibility has to lead somewhere and in the case of this book there’s one great example after another from design best practice, the history of visualization (who knew Florence Nightingale was a DataViz practitioner) and just great visualizations (as well as bad ones for ‘what not to do’). Below are a few of my favorites:
First up it’s the classic, the chart (technically an Infographic) often stated as being ‘the greatest’ and I largely agree; Charles J Minard’s map of Napoleons march to and from Moscow in 1812/13:
Whilst this is an Infographic and not a chart per-se (An Infographic is tailored to the dataset where as a chart / data visualization can be re-used against more than one dataset as in Marey’s chart below.) it is a perfect example of the potential power of what we do, it shows clearly and simply how the French Army was decimated by the Russian Winter as well as the Russian defenders.
Secondly I have to mention EJ Marey’s Paris to Lyon train timetable chart as it’s a great example of data visualization despite it being created over 100yrs ago –it also graces the front of Tufte’s book:
Take the time to interpret what’s being shown here; an entire train timetable in one chart – brilliant.
That’s enough of the examples (there are plenty more from the book) as they’re not Tufte’s work so he shouldn’t be getting too much credit for the work of others. Where he does deserve a great amount of credit is when he covers chart best practice in a clear and easy to follow manner ideal for the novice. Along with Stephen Few he’s achieved great things in the Qlikview space in getting visualization novices to adhere to at least the basic principals and common sense of good design – there’s still a looonnnngggg way to go judging by the majority of Qlikview dashboards out there. Personally when I read the book for the first time I found that it was more of a reassurance – ‘you’re already doing the right thing’ as opposed to a guide to new practices as I’ve always taken the time to design things properly. Indeed many readers may find that much of what he recommends (the data to ink ratio for example) is nothing more than common sense but it’s no less valid and essential knowledge for those who don’t know it.
I do have to take a slight issue with some of the things he says however. For me, in much the way as Stephen Few, Edward Tufte takes things too far by recommending that too much is removed from a chart and thus ease of viewing and engagement are lost. For example in a section covering how to effectively display a bar chart he gives the following example:
Here he’s recommended removing the gridlines and replacing them with a background coloured grid over the top – and thus reducing the data to ink ratio. For me this goes too far; the chart now looks as though it’s a stacked bar chart. That’s just one example of the general issue and I don’t want to seem like a graphical pedant as many armchair data visualization ‘experts’ seem to be; I would much rather have a dashboard or chart built to Tufte’s / Few’s principals than one that isn’t.
So other than a minor (an it really is minor) disagreement about the extent of data to ink ratio’s I’m fully on-board with the recommendations espoused in The Visual Display of Quantitative Information; I love the examples, the history and even the way the book is printed (it’s published by Tufte himself and that really comes across in the style), it really is a must for anyone working with Qlikview or data visuals in general.
Summary: Great book, largely good principals and an ideal starting point for anyone in the data visuals business.
Qlikview Rating: 5/5 Overall Rating 5/5
In the course of writing this review I’ve had a hunt around Amazon and found a copy of Marey’s ‘La Methode Graphique Dans Les Sciences Experimentales Et Principalement En Physiologie Et En Medecine’ first published in 1885: http://www.amazon.co.uk/Graphique-Sciences-Experimentales-Principalement-Physiologie/dp/1143726219/ref=sr_1_3?s=books&ie=UTF8&qid=1334319964&sr=1-3 My copy is on order although I have no idea how I’ll translate it (page by page via Google image translation??).
However there’s a final crucial point before the end of this review…
This isn’t specifically related to the book so the above review still stands, it’s to do with how you use the book (The same goes for Stephen Few’s: ‘Information Dashboard Design’). PLEASE, PLEASE read the whole book (or at least the relevant sections) and THINK about what it’s telling you, flicking through the book and looking at the pretty pictures and taking a few sound bites away doesn’t make you an instant visualization expert. I come across so many people and blog or forum posts exalting ‘Stephen Few wouldn’t do that…’ or ‘I’ve made this according to Tufte’s principals’ where the principals have been applied either in the extreme or in a half-baked method that helps no-one; blindly removing axis & annotations can be disastrous. Both Few & Tufte state that their general principals should be applied ‘within reason’ – a key phrase that many readers of their work fail to appreciate. Many of the things recommended in both books should be used as good guiding principals and not all too easy to follow rules (‘maximise the data to ink ratio’) you should always use your own common sense and judgment and ask yourself the simplest of questions ‘Is the chart I’ve created as easy to read and interpret as it could be?’ – if it isn’t change it. Always look at your completed work from the users point of view not your own developers stand point (The issues & dangers of ‘Dashboard Developer Myopia’ will probably be in a dedicated upcoming post).
As always I hope you find this useful.
Happy reading & all the best,