Effective data visualisation for the WIPUP dashboard.
Data visualisation is a big, big topic. It’s often underestimated by the majority of society. But what is data visualisation? It’s not so much how effectively you present data – no, it’s how effectively you deliver information. Today since work has resumed on WIPUP I decided to take a look at how to implement effective data visualisation in the WIPUP dashboard. Before we continue, let’s take a look at how the dashboard looks like now:
Obviously unfinished, the dashboard looks like somebody vomited. Disregarding the obvious visual glitches such as the 0% segments in the pie chart and questionable use of colours, the most obvious thing that immediately catches our eye is … nothing. That’s right. It’s a mosaic of data, no clear correlations or focus is delivered to the user. Well, what would you expect if 6 charts were thrown at you? It’s not hard to see that there is a problem here to be fixed.
The first step is to consider, based on the results of dogfooding, which graphs aren’t realy useful to the user at all. The two obvious ones are the comment activity and the popularity based on subscribers. If, hypothetically, there was a popular WIPUP user, the usefulness of these two will increase, but overall it’s quite agreed that the majority of users will not find it useful, and even if showing a variety of data, we have to look at the purpose behind comments – to initiate discussion. Such a qualitative concept is not well presented with quantiative means and thus should not be attempted. Looking at the subscribers chart, it may be useful, but doesn’t warrant its own graph.
The second step is to consider exactly what medium we are using to present the data – in this case we are using line charts and pie charts. Line charts are appropriate here as it is used with a timescale, but as for the pie charts – well, experts unanimously agree that there is absolutely no usecase for piecharts. Pie charts should not have been invented. They are useless at showing data. It is hard to determine which segment you should focus on, and the colours distract rather than emphasise the data. Even worse, I have commited the ultimate evil of creating a 3D pie chart. Here’s something you should try out – cover all the percentages of the “activity per project” pie chart, and look at the two large segments (purple and green). Looks about the same proportion, eh? Nope. There’s a stunning 10% difference between them. Try the same with the orange and yellow together – big relative difference again. This trick even works with groups – the orange and yellow segments together look around the same as the pink and lime green, no? Yep, again, there’s a big difference. Conclusion: piecharts suck. The only benefit is that they look pretty.
The third step is to search for correlations. The most obvious one is how many updates I have made and how many views I’ve received (duh). The important thing here is that from the user perspective, they want that correlation. They want to see that the more they update, the more views they get. Thus we should try and emphasize this correlation. How better than to merge the two graphs? This way any mismatch of this ideal correlation (ie, an update they thought would be interesting, but from the statistics, show otherwise) would be obvious.
After a quick brainstorm, I copied the raw data onto OpenOffice Calc (equivalent of Microsoft Excel), and tried this alternative:
As you can see, I’ve compressed the data of 5 graphs into 2 graphs (say, I didn’t know Calc could do 2 y-axes). On the left, we see a stacked bar chart showing activity, kudos and subscribers (trackers are added as a subscriber to all projects, which is why it is all equal). Immediately the user can create a focus and see correlations. For example, this allows us to sort the data in terms of most to least. In this case, I’ve spent the most of my updates on uncategorised updates. However, we can now easily determine the correlation that even though the activity spent on uncategorised work is the most, it doesn’t create as big an impact as say, the Evan project, of which it is clear that the Kudos+Subscribers:Activity ratio is much greater. Infinitely more useful, don’t you think?
On the right, we can see we’ve showed the correlation by merging the two line graphs. Now we can clearly see that there is a slight correlation misalignment in the update I did on the 07/06 – perhaps the update wasn’t very interesting? Or perhaps it’s simply because I wasn’t as actively giving out updates as before – all very interesting conclusions.
Well folks, I hope I shared some useful stuff today. Next step, implementing the changes and revamping the dashboard.
No related posts.
Erik Kylen says: (16 June 2010)
3 Comments