10/09/2014

This is America !!

IronViz - Part 2: Contest and Viz


This second post is here to relate you my IronViz in Seattle. Already one month and I can tell this was a dream !! Indeed, 3 weeks before the #DATA14, I would not able to predict this journey. That was an unbelievable adventure. For me and my wife (she accepts to be with me in a conference linked to my job !) it was my first time to the United States of America !! And it was amazing ! Each person I met in Seattle was gentle and lovely. I can espacially thank Tara Walker, Jewel Loree and Ross Perez from Tableau Software, and Jeffrey Shaffer and John Mathis my contestants for their welcome. They were very gentle with me in spite of the language !


But let me go through my topic: how I lived the IronViz contest ! First, it was very impressive to be a part of a huge conference like #DATA14. It was funny to see my biopic in the section "speakers" on the TC14 website... Around 6,000 people included Tableau Zen Masters and other VIP. It was the first time for me to be in a Conference like that. I was present in TCEU 2012 (Barcelona) and 2013 (London) but in Seattle.... Waooh !! Many and many sessions, speakers, sponsors, Tableau-junky... With so many people I just was able to attend keynotes. On the other hand, I have to admit that the IronViz contest was in my mind during all the conference... The contest, that was the biggest thing in my mind during this week. The rules were clear: 20 minutes, 1 dataset, 3 contestants and 1 winner! I was very nervous...

The viz is at the end of the post.


1. The dataset & the idea


The dataset was about Yelp, a search engine specialized in restaurants, casinos, hotels... The data included Las Vegas and Phoenix (in fact they included more data because of some mess in the file...). There was a lot of lines in the extract (more than 1 million) so the workbook was a little bit slow during opening and during analysis. To analyze the dataset and find a good idea for the contest I used the same process than I use with customers: discovery, idea, creation, user experience. This process allow me to do several iteration and try to improve the viz. The delay was very short so the number of iteration too. But thanks to this process I was able to find new ideas during each iteration. For example, I find the idea of totals during the last one. Each iteration was a training for the contest so I tried to do quicker each time. To be ready the D-Day, it was important to know each step of the creation by heart. I can tell that all every night was short !!

My idea was to tell a story to the judges. This story was about notes attributed to business locations. Because of the dataset's size I chose to focus my viz on Las Vegas. That is why I put some global filters to limit the perimeter and analyze Las Vegas'data only. The first graphe I built was the map. In my mind I told me: "If there is geographical coordinates, try to do a map !" I think it is the better way to quickly see each location. All the graphe I built for the viz have the same color legend based on the number of stars attributed to places. It is important because it was like a road to follow during the story. Because there was a date field I chose to create a timeline with the number of reviews for each number of stars. The aim of this graphe was to show that most of the notes are between 3 and 4.5 stars. Two conclusions came to me: reviewers are very gentle with places or only the reviewers which love place write comments... The last two grahes are lists about places. One shows the top/bottom 10 of places in Las Vegas (in terms of number of reviews) and the sceond one shows the total of places for each number of stars (this one is complementary to the timeline.


2. The D-Day


The IronViz day was the thursday on the morning (the night before was very short...). I was very nervous to do a viz in less than 20 minutes in front of thousand attendees. Hopefully my sous-vizzers, Tara Walker, helped me during all the week to be ready the D-Day. With all the training I did during the week I was confident about the time. On the other side, I was more sceptical about my skills in statistics. Indeed I bet all the contest on the storypoints and the easy of use. That is why my viz does not have complicated graphes and consequently strong analysis.

The process to build tmy IronViz is simple: 3 dashboards with 5 graphes each. Each graphe is build then the 3 dashboards. In chronological order, I first put the filters to limit the number of lines. Then I built the map with Latitude and Longitude and put the number of stars in color. After that, I built the top 10 sheets with the creation of an index. To build this sheet I chose to shut down the automatic updates to avoid wasting time and I put all the dimensions needed and the index. To create the top I configure the index to be calculated in a descendant way based on the number of reviews. Of course to create the bottom the index was calculated in an ascendant way based on the same measure. The mark type is shape and was configured to display stars. The timeline was built with the review date and the number of reviews. The dimension number of stars was put on colors. The last graphe was the total built with the dimension number of stars and the measure number of places (it is a distinct count of ID_business).

To create the story it was necessary to build 3 different dashboards. Indeed, the first one displays all the graphes with the 10 worst places, in the second one the 10 worst places is replaced by the top 10 and the last one displays all the graphes but have 3 quick filters allowing the users to seek for places. The storypoints allowed me to exclude especially values of the dimension number of stars: the first point excludes data with a note higher to 2 stars and the second one excludes data lower to 4 stars. I succeeded to do the entire viz in 17 minutes. I was proud to be so fast... but after the contest I told me that it might be better to do it slower but with more statistical points.

The judges were very kind with me and my job. They told me that my viz was the more simple of the contest. The colors was a good choice because the red is related to bad stuffs and the green is the opposite. They loved the storypoints and especially the focus on the worst places because in most of cases the focus is done on the best things. In the final viz, I put quick filters with drop-down menus. The judges told me that it is not the best way to allow users to look for a specific place. That is why, in this post, I replace menus by typing boxes.

You can find the viz at the end of the post. Please comment and share, I will be glad to discuss about it !


Feedback


After one month, I can tell that it was very rewarding to do the contest. Right after the contest I have the chance to met Ryan Sleeper the 2013 IronViz Champion, Kelly Martin a 2013 IronViz contestant and now Zen Master, Ramon Martinez, Jock Mackinlay... and other impressive Tableau users.

Of course, congratulations to my two opponents during the contest Jeffrey Shaffer and John Mathis ! You can take a look to their IronViz feedback here for Jeffrey and here for John.


1 comment: