How I Came to Choose One out of Sixty Two Data Visualizations

Disclaimer: This is an unofficial post about #IronViz contest held by Tableau. I have written about my adventure on how I chose my final vote for 1 out of 62 entries and how the process of chosing one taught me a lot about better data visualiztions.

terrific entries for geospacial #ironviz contest. Image: Tableau Inc.

terrific entries for geospacial #ironviz contest. Image: Tableau Inc.

GeoSpatial #IronViz 2017

Mesmerized, inspired, joyed and educated! Huge thanks and congratulations to amazing #IronViz contestants.

Voting for Tableau's GeoSpatial #IronViz closing in another 20 hours and I am yet to review and cast my vote for the best of best entry out of 62 submissions. Certainly not an easy task.

Evaluating the work of 62 amazing designers and analysts; including some Zen Masters, made from 62 different data-sets. Being honest, objective and fearless was certainly not going to be easy.

As I opened the official entry-page and saw the list of all 62 entries, the first thing came into mind,"Is the order of all the entries is random?" To confirm, I refreshed the page (to see if the order will change), to my surprise, the order didn't change. Will loaded full of all the previous biases like my personal preferences for data visualization style and color, the kind of visualizations I make and how personally I know authors of visualizations, I could not afford another bias for the order of appearance in which these entries appear. I quickly wrote a python script and reshuffled the order in which I would evaluate visualizations:

# number of contestants = 62  
contestants = 62  
import random  
entry = list(range(contestants))
entry = [x+1 for x in entry]
random.shuffle(entry)
print(entry)

Criteria to Evaluate Entries

Tableau defined four criteria to evaluate entries

  • Design
  • Storytelling
  • Analysis
  • Overall Appeal

I reordered list of entries and started evaluating the entries. The first entry I came across was by Luke Stanke - Impressions of Climate Change (last week itself I was listening to the post-cast by Enrico Bertini and Moritz Stefaner on Are animated maps batter or worse than multiple static snapshots of maps to visualize geo-spatial data in temporal dimension) Beautiful visualization! I knew there was plenty in store.

Methodology

I evaluated each visualization in each of four parameter (design, storytelling, analysis and overall appeal) on a scale of 1 to 5. After I evaluated ratings to each parameter, I took mean of all four parameters and assigned the mean as the final score to that visualization.

Example: I rated Anya A'Hearn's amazing viz as following on all four parameters:

Design 5
Storytelling 5
Analysis 3
Overall Appeal 4

Thus I assigned final score of 4.25 out of 5.00 (average of 5, 5, 3, 4) to her entry.

When setting up machenism to rate entries, Initially I was tempted to assign more weight to overall appeal than other three parameters, but I dropped the idea first, because it was nowhere mentioned on Tableau blog and secondly, it would underweight other important aspects if my personal judgment of overall appeal heavily relied on any of one parameter.

A Really Tough Choice to Make

It is easy to judge than being judge, I know everyone has put a lot of hard work and efforts into making these entries. Almost all of entries were amazing, some impressed me more and some less. Finally there were 10 entries that were incredibly good to not to be voted for.

I had to have a look again. And following five entries ended up being in the top of my list (in no particular order)

  • Anya A'Hearn's Transposing War
  • Joshua Milligan's The Changing Shape of History: From Colonies to the United States
  • Timothy Vermeiren's The History of the World
  • Dustin Cabral's Chasing the Sun
  • Marc Soares's Invisible Walls: The Reality of Racial Segregation in America

Anya got two perfect (design, storytelling), Timothy got two as well (storytelling, overall appeal), Joshua and Dustin both earned three perfect scores (Joshua - Storytelling, analysis and overall appeal, Dustin - Design, storytelling and overall appeal) and Marc had 2 perfects for storytelling and design.

A tie between Joshua and Dustin! Each scored 4.75. I spent another 20 minutes looking at those entries and finally cast my vote in favour of Dustin Cabral (Not being able to vote for the superb entry of Joshua Milligan kinda killed my heart) for it's emotional appeal and winning my immediate trust into the key message he tried to convey.

Other Entries Definately Worth A Mention

Jacob Olsufka's Hawaiian Island Rainfall for design
Olga Tsubiks's Mapping out the Unknown for design
Marc Soares's Invisible Walls: The Reality of Racial Segregation in America for storytelling
Neil Lord's To Pret or not to Pret for storytelling
Emily Chen's 2016 Math Results in New York Public Schools for analysis
Pablo Saenz de Tejada's Madrid In Detail for analysis

The List of Contestants with Scores I assigned

I thought a lot about making my scores public. I am no official Tableau employee, I am no Tableau authority, just another kid on the block, my ratings might have a lot subjective factors, I have friends in the list whom I judged, I am nowhere close to contestents in the list. But then I realized, when provided creatively and with all good intentions, everyone loves the feedback. And more than that, the list belong to the community more than it belongs to me, So here it is:

I am sorry, if I might have hurt anyone.

Few Words to Tableau

Thank you Tableau for putting the amazing show. Contests like these are not only engaging but very educative and inspirational for a lot of people out there. I learned a lot from these visualizations and will sure dig deeper in some of those. You know the kind of positive feeling when we discuss these incredible visualizations with our collegues and friends and we pop our mouths open in the awe! And the very next thing we do is to download and try to recreate the visualization. This all is possible because of tremendous resources Tableau puts into Tableau Public.



And finally, we are all humans and it is extremely easy to nudge us.

t-test, anyone?

t-test, anyone?

With all due respect, I would like to add, "If possible, for all future contests, it would be amazing if the list of contestants can be randomised for each pageview. Afterall we are nerds ;) and would make any improvement that is possible with data."

Closing Remarks

Thanks to all contestants for taeching me a lot about better visualizations and maps. Cheers to everyone.

Your friend, Ashish Singh