Recovery Review

Visit Recovery Review

As I was adding links to this blog, I went to the Recovery.gov download center and discovered that the stimulus data was updated about a week after I started working on my project.

This doesn’t really affect my project in terms of the visualization or the architecture, but I ran a couple of tests and it looks like the specifics of some of the stimulus projects is different enough that I will need to revisit the concept of flags in the data so that the flags are “release specific”. This is because it looks like some of the problems with the data are fixed with subsequent updates.

Another thing I would like to do eventually would be some way to double check one version of stimulus data against another version. For example, I noticed that there are substantial differences between some of the projects in the 2009 Q4 data release and the 2010 Q1 data release. In one project that I looked at, the difference was to the tune of $600 million!

This obviously requires additional investigation, but I’ve put everything on hold until the Sunlight Foundation contest is over.

This tutorial walks through the process of adding a link to give more context to a stimulus project. In this tutorial, we link to the Wall Street Journal blog post on a piece of inaccurate data (which has since been fixed). Adding links helps users see the stories behind the project that the raw data represents.

This tutorial walks through the process of adding a flag to indicate inaccurate data in a stimulus project. Adding flags helps others quickly see which projects have data that is false, misleading or otherwise unhelpful.

This video is a quick (8 minute) overview of the functionality and features of Recovery Review.

I think I still have a couple hours before the Sunlight Foundation’s Design for America contest officially closes. Also, for anyone from the Sunlight Foundation, please take this blog post as evidence that I believe that I’ve submitted my project. It’s really hard to tell if I actually did it right, but I’m assuming that I made it.

I wanted to address a couple things  that I either didn’t get around to implementing or various design decisions I made along the way to create Recovery Review.

Silverlight

First of all, the decision to make it in Silverlight. The answer is simple: Because Silverlight is what I’m best at. In my normal life, I’m an interaction and UI designer and I’ve enough experience with Silverlight that I’m reasonably fast at it. It also has a pretty robust set of tools and services and is great for making really compelling interactions and experiences. I’ve done a number of visualizations using Silverlight and it is my go-to technology.

Mapping the Data

I have a soft spot in my heart for mapping and geocoding things. It’s fun. It’s interesting. It’s an easy way to pull people into a project because we all are from some place or another. We see something we recognize and it immediately resonates with us. So I decided early on that I wanted to take the stimulus projects and so SOMETHING in terms of mapping them. I chose the Bing Maps SDK because it takes about 15 minutes to set up and start working in Silverlight.

The problem with the map was that the data became kind of obscured. In a given search, there were several projects located in the exact same place… usually at a State Capitol building or within 10 miles of it. So I had to come up with a solution for showing multiple items at the same latitude/longitude coordinates. I did, but I was never really happy with the results.

I had to have the items semi-transparent so that they could be seen when overlapping, but that just ended up requiring an enormous amount of maneuvering on the part of the user. Overall, I didn’t feel like I was getting the experience bang for the time buck that I was putting into this particular visualization.

Fortunately, Silverlight is pretty robust and I was able to keep what I did have, but just push it back into the corner of the application. It’s still interesting information, but it’s not the star of the project.

Deep Linking

Oh, how I want to do this. One of the things that I was aiming for when I started this project was the ability to let people link directly to a search query or individual item. This is possible in Silverlight but it managed to wander down on my list of things to do as the deadline grew near. I’m taking a week off of development starting now while I wait for the Sunlight Foundation judges to make their decision, but I hope to work on this when I’m rested up.

More Social Web Friendly

I originally had it so that users could add far more information to flags and links that I ended up with. I wanted easy ways to connect this information to Facebook, Twitter, MySpace, etc, etc, blah blah blah. The problem was that the flag and link screens started looking like form soup with more than about 4 fields in them. I wanted something that was light and fast to use, so I kicked out the non-essential form fields in the interest of simplicity. Plenty of things that could have been implemented weren’t in the interest of making the core experience light and fun. But that came at the cost of things like…

Logging In

Again, it was in my original plans and I have much of the framework set up to do this so that users wouldn’t have to add e-mail and name every time they wanted to do something. But as the project deadline loomed above me, this got kicked to the side. I’d still like to put it back in at some point.

Zooming In and Out on the Visualization

This design may look simple, but it took many iterations of design to come up with something I could smile at. Believe it or not, most of my failed attempts were much more complex than this. Which if part of the reason they ended up getting discarded. I wanted something that was simple enough to just understand and could become second nature the second or third time the user sees it.

Above is the resulting visual of a search for all items costing more than half a billion dollars. (Click to see the bigger version.) It gives a pretty good understanding of how those items fit into the award and job reporting whole. But when we click on one of the items, the larger circle becomes huge so that we can see the individual item in a scale that is more conducive to visual understanding.

And that’s the biggest stimulus item in the database. If we look at something “small” like a $50,000,000 project that reported 625 jobs, it looks like this:

In relation to the $182 billion listed in the database, a $50 million project is so small we can hardly detect the curve of the line. However, with this data set, this is, I believe the best we can do. We simply have such a huge range of values that we have to make a compromise somewhere. I’m only sad that I didn’t manage to get some kind of zooming so that we could interactively see the total funding in a more digestible size and thereby see how tiny $50 million can be in relation to it.

Well… that’s all I can think of at the moment. I’d love to talk more about the things that I like about the project (there are tons of them), but it’s often more instructive for me to learn from the things I didn’t get right.