The Art Market for Dummies: Building a Data-Driven Explainer


By Jean Abbiateci, French freelance journalist and teacher at Lannion School of Journalism.

Art Market for Dummies is a data visualisation I made last spring, and one of the winners of this year's Data Journalism Awards.

The visualisation tells the story of a topic that is little known outside expert circles - the international art market - and that I myself would have liked to see as a reader. Everyone knows Picasso or Warhol, but few people know that art represents a bigger market than the cinema and the music industry. I am not an expert on this topic myself, so I wanted to create an application that enables users like me to understand the subject quickly and easily, in three minutes. And so I embarked on the creation of this data journalism piece, together with the French publisher Askmedia.

Art Market for Dummies

Many people associate data journalism with serious “hard news” data such as election results, census data, and unemployment rates. I wanted to work with a lighter topic that can surprise the reader (as does data explorer Jon Millward with his porn market database). Of course, I would be delighted to work on a very ambitious data-driven investigation such as ProPublica’s Dollar for Docs, but, when it comes to small newsrooms and freelance journalists like myself, I believe small unconventional projects are the future of data journalism. 

I had already done a visualisation on the topic a year back and was surprised by the large audience it generated. I knew, as a result, that the blend between art and money interested the audience.

When developing the project, I tried to focus on its pedagogical aspects. I once read that journalists must be like guides in a museum. I believe this is an important lesson for data journalists, especially since data visualisations are sometimes beautiful without being easily understandable to readers. I was greatly inspired in this sense by the advice provided by Studio 20´s project, Building a Better Explainer Project, launched in November 2010 by professor Jay Rosen of NYU and ProPublica.

The data for my project comes from various online databases (such as Artprice) as well as press kits. I used the Outwit plugin to extract the data from webpages. This software is absolutely awesome for beginners. I have some notion of programming but I would rather do without the extra work… 

One of the main issues I faced was that the value of artworks was expressed in different currencies. To address this problem, I figured out a way to grab the value of a currency on the day of the auction using the Google Finance API.

On the technical side, I reused many elements I discovered on GitHub (via the Source project): D3.js and Isotope.js for my timeline; a project from MinnPost for the menu; and a great piece of code from Jim Valhingham for the bubble charts. This was the first time I relied so much on GitHub. All these repositories were very well documented and it was fairly easy to adapt the code.

The most important takeaway from this project is the necessity to tell stories with data. One of my visualisations has a series of filters that enables users to sort out the 320 most expensive artworks sold around the world using various criteria (such as nationality of the artist, location of the sale, etc.) The most popular filter by far is the one enabling readers to visualise the gender distribution in the art market. This filter shows that only one woman appears as author among the top 320 of the most bankable artworks in the world. Many people have tweeted this link and said: "It's outrageous, it's so macho!”

Gender distribution visualisation, Art Market for Dummies

I would have liked to find more filters that are able to tell a story as good as this one, but it's not easy. I wanted to know, for example, what the dominant colour in best selling artworks was, in order to answer the question of which colour enables an artist to make more money. I tried using the Image Color Summarizer API, but the result wasn’t very convincing.

To sum up, I learned quite a lot from this project, including from mistakes I made. One of the lessons I learned is not to try to work on the data analysis and on the graphics at the same time. It is very tempting to think: “I am going to create a graphically striking data visualisation", before doing any analyses, but this is a mistake. The process of creating a story in data journalism is identical to traditional journalism: Do not start writing an article before you find a good story. Analyse the data with Pivot Tables in Excel first, to see what story the data can tell. Only then start thinking about what visualisations can be designed. This is the best way to work efficiently without losing time.