Oh man guys, I had some fun with this one. What you are about to feast your eyes on is data scraped from Etsy using IFTTT, which is quickly becoming my favorite set-it-and-forget-it data tool. I scraped anything tagged “Hello Kitty” from August 14th to September 17th.
I didn’t expect any crazy revelations in the data besides that there is A LOT of Hello Kitty crap on Etsy. And most of it is under $5 it seems! I really wanted to do this because I knew it would be a blast to design. And oh man it was.
I followed the instructions to add colors to my preferences file. Originally, I had this in as a “regular” palette, meaning that it was meant for discrete data. But, it looks pretty continuous to me, and all my data was continuous, so I made it an ordered diverging palette! Weee! Isn’t this fun already?!
Next, I wanted the text to be just as adorable as the colors. I switched pretty much everything to Comic Sans (I know… boo! hiss! but John Maeda said it was ok at his TCC13 Keynote!). I also downloaded a sweet font off of dafont.com called “Loveness Three”
I grabbed a vector image of Hello Kitty and blew it up really big to make the header. I felt it was a particular stroke of genius to put the hover help menu (my signature trick!) in her bow. How nifty!
I’ll go into more detail about the design work behind this viz next month, since it will be Design month at Tableau Public! Without further ado, here’s a dashboard with more hot pink than you will ever see on Tableau Public ever again:
Which led to a viz that at the time looked like this:
The original viz was created by a mysterious group (for which I have no contact information so please stop asking!) called the “Pornhub Stats Division.” What you are looking at is the average duration on the website Pornhub and the top 3 searches in each state.The data was so intriguing, but it was apparent that whoever made this was new to Tableau and was missing some of the finishing touches. I wanted to give it a quick facelift so that it the information would be more digestable, so I did. I liked what they did with the formatting of the bar chart on the original. It matched their logo and it really popped. So, I went ahead and changed the map to a dark map (From the menu: Map>Map Options>Style>Dark) and changed the dashboard formatting to a black background (Dashboard>Format>Shading>Default). I also got rid of the table at the side since I didn’t feel it added much. I used duration to color the states and had the top search display as text. I also cleaned up the tool tips a bit. If I had been feeling like spending a little more time on it, I would’ve floated Alaska and Hawaii under the U.S. so that you could easily see them too. If I had felt like spending a lot more time, I would’ve reformatted the data so that you could click on a search term and see all the states that had that in their top 3 searches. But, it felt kind of weird to have it say certain words all over my monitor at work. Anyway, this is how it turned out.
The tweets were (and still are) pouring in. I was getting random emails from friends and colleagues about the viz. It started gaining thousands of views an hour. I permalinked the original behind mine, as to not steal the original author’s thunder. Now both vizzes are pushing 300k views. It was definitely the top viewed viz last week.
A couple of thoughts:
I was interested in how most of the discussion was about “What do people like in MY state?” Not only does the internet love porn, it’s also fairly narcissistic, it seems.
I was surprised with how many times I had to tell people to hover over Utah to see what they are into. The idea of interacting with a viz is still not intuitive to people.
A lot of people wondered how I felt being associated with Pornhub. To me, data is data and interesting data is interesting. And really, nothing makes me more sad than an interesting story being covered up by poor design. All I set out to do was make the pornmap readable. They incorporated some of my changes and got hundreds of times more views to that data, so I think I accomplished that.
If I was Pornhub, this is how I would follow up:
Data on average visit duration based on search. Do people who watch certain things last longer? I’d add variance as well, so we could see if there is a wider range for some topics.
Reformat the data in the original viz so that you could click on a search term and see how many states have it in their top 3. They had the keywords spread across 3 columns: 1st, 2nd, and 3rd. Instead all of the words should be listed under one column with the rank listed in a separate column. There would be three rows for each state. This is pretty much a perfect use case of the Tableau Reshaping Tool Excel Add-in.
Add variance to the duration times in the original viz. There were lots of complaints that the data isn’t really telling you much, which it isn’t.
Worldwide data would be a no-brainer.
Mapping where uploads come from and where the most active users are.
I’d love to see if there was an increase in some of the more specialized searches like Anita Queen and Smoking. Google Trends suggests that there might be.
In closing, what did last week teach us, boys and girls? As the good folks at Avenue Q so eloquently put it:
I’ve wanted to make an Instagram viz since I started using Tableau. With our Social Media Viz Contest launching, it seemed like the perfect opportunity to get something together.
Paul Gambill helped me by writing a script that sent calls to the Instagram API to get back a JSON file filled with information about posting times, content, captions, comments, likes, and my favorite piece of data: what filter was used. The method he used pulled posts that were tagged with the tag of my choosing: #nofilter.
A little backstory on my obsession with the #nofilter tag:
I’ve been wanting to make a Phish dashboard for a while. I wanted to use the phish.net API and collect all sorts of delicious, raw data of setlists and locations and create all kinds of things. But alas, it doesn’t look like I will be allowed to. I gave up for a while, but then this tweet popped up in my feed:
I took the data, which looks identical to phish.net’s song list, and put my spin on it. Not quite what I was hoping for, but I think it’s kinda neat. I do appreciate that the “Cyclic” color theme almost matches the Phish logo perfectly.
Sasquatch! Music Festival has a killer lineup this year. It’s no surprise that tickets sold out in a record 90 minutes. With tickets going that quickly, people are turning to the after market to try to score wristbands. All over the forums, ticketless fans were repeatedly told to wait until a couple of weeks before the festival, that way ticket prices will be close to face value. These graphs were floated around a lot:
The pattern this year, however, is a little absurd. I’ve been scraping the Seattle Craigslist for posts and I put the data in a Tableau viz. As you can see, ticket prices have not had any kind of consistent downward trend. They are rarely offered for below $400, even now that we are only weeks away from the Memorial Day Weekend festival. In my personal experience, half of these posts seem like scammers, too. One in particular keeps saying that they will only deal over eBay and asks for a bunch of personal information so they can “set up the transaction.” I bought my tickets presale, so I’m all set to go (thank god!), but I’ve been looking for tickets for my friends and it has been pretty rough. I’m not sure if it’s just increased demand that is leading to ridiculous after-market prices on Sasquatch tickets or if there are other factors, but I think we can all agree that these kinds of scalpers totally suck.
UPDATE 5/21/2013: I’ve been updating this viz basically everyday which involves cleaning up the dataset. There’s only 3 more days until Sasquatch and judging by the amount of “WANTED!” posts I had to delete from my data, I’d say it’s definitely a Seller’s market right now. Perhaps the age old wisdom of “Wait until the last week and prices will go under face value!” is proving to be obsolete.
I was once again exploring the last.fm playground when I viewed the following graph. The graph charts my top 90 artists of the last 12 months based on the average age and gender of its listeners. Apparently, I listen to music that is very much in my age group. Something I found interesting was the gender split. It really seems to correlate with the two genres I primarily listen to: more folksy things tend to have more female listeners and more electronic things seem to have more male listeners.
For on and off the past year or so, I’ve become interested with the Quantified Self movement. The idea is that if collect data about ourselves and analyze it, we can make inferences about our behavior and adjust accordingly. I’ve been using a few iPhone apps to track different things that I’m interested in, but it’s struck me how inconsistent I’ve been about keeping up with them. Recently, someone posted on Reddit a link to a years worth of their sleep data, wondering what to do with it. I pulled it into Tableau and played with it a bit, which was fun. My favorite view I came up with was this heat map comparing the hour that the person went to bed versus the hour they woke up. The size is the number of times that combination was recorded and the color is how many hours of sleep he averaged (green->blue=short->long).
I liked that view and thought it’d be cool to explore. I have also been collecting sleep data for a little over a year, so I thought I’d see what mine looks like:
Kind of sad. I have data dating back to February 2012, but I’ve been pretty inconsistent with collecting the data. So, there wasn’t a whole lot to look at. That and I have a job with very steady hours (especially compared to the schedule of someone who is a student like the person who provided the data for the other heat map), so I go to bed and wake up at pretty much the same hours every day. Not all that exciting, huh. So, although I couldn’t really find anything good is my incomplete sleep data set, it’s at least encouraged me to be a little more consistent about my data collection. Maybe in a couple months I’ll actually have something worth mining!
BitCoin was all over the news today, hitting a historic high of around $230 this morning. Since then, it’s dropped back down to around $170 at the time of this post. But let’s be real people, I remember jumping on the BitCoin train when it was $1, so either way, this is pretty cool. This morning I scraped the website BlockChain.Info to get some BitCoin data. I quickly through together the dashboard below. It’s sort of a “Choose Your Own Adventure” style dashboard; I used parameters to allow you to choose what you want to see visualized. I also made my dataset public, which I’ll try to update with the most current numbers sometime soon.