Information Dashboard Design, 2nd Edition

One thing I've observed is that a user's requirements always flip, and they're usually not aware of it. The two biggest bits of feedback I've heard are "too much data" and "not enough data" -- often from the same person, and occasionally on the same dashboard.

People just want the highlights, except when they don't, in which case they want everything. The former is really hard because highlights require a lot of rule-based context. The latter is also hard because "everything" is often meaningless. Too much flexibility in your data exploration tool and people just go on random fishing expeditions.

Now I try and structure data in terms of a conversation, with an information "needs" to determine the priority order. Start with the hurdle requirements (e.g. something key like sample size), move through the highlights then present the bare minimum to guide further exploration. If you make data exploration easy from that point it this seems to work well, but I'm still learning.

I must admit, I took one glance at the "ideal dashboard" and was a bit bewildered. I had no idea what I was meant to be looking at, or the relative importance of things. Perhaps there is some critical, specific domain knowledge that I'm missing. Either way, will definitely get this book to find out more.

What's worst is when offering free trials of an analytics product and then realizing that all freemium customers engage with a dashboard once, and then dropoff forever. Event tracking and various stages of a/b testing can be dangerous because they steer engineers to local minimums in designing a product.

Having a logical understanding of every element of the dashboard is the best way to approach the design.

Great post, we need more resources like this out there.

In this case, I was super-surprised that the letter grades were barely visible on the winning design. The post-competition design was far better in the respect that it displayed the key takeaways prominently.

The register of attendance was lacking in several aspects compared to the common/traditional way it's done on paper. For example, since we probably need to see exactly what dates/days/lessons were missed, having the data in a standard per-week format would be more useful.

From an educational point of view, we should want to use computers to improve the quality of the data, rather than just displaying the data prettily. It would be better to link together the available information to provide educational insight.

Ideally, we might want to say something like: "You missed the class on X, and performed poorly in those questions on the associated test, so you need to catch up by studying the following", using the dashboard to access this type of info.

Further thought: It would be awful to have the data displayed worst-to-best. The comments pick out that this is because teachers should spend more time on the worst performing students - in my mind, this is a fallacy (under-performing students, perhaps yes). Worse than that - I would typically want to easily identify the data for a particular student, in which case the data would be better in some consistent (alphabetical) order.

For more exploratory information design, check out the stuff by Nicholas Felton. He's a big quantified-self guy who every year publishes a really nicely designed annual report of his life [2]. I emailed him once asking for more sources on info design and he linked me to the site of a course he taught that contained some great resources [3].

[1] http://www.amazon.com/gp/product/0195135326/

[2] http://feltron.com/

[3] http://interactiondesign.sva.edu/classes/spring11/informatio...

The main focus has been on readability from long distances and the ability to switch context when up close using the leap motion.

Check it out here - https://www.youtube.com/watch?v=kMM2rPX2Rok

There seems to be some pretty useful information in this book that we can apply to our current setup.

Dashing here - https://github.com/Shopify/dashing

Thanks!