I've just had a good look at Luminocity 3D by Duncan Smith of CASA, Bartlett UCL. I'm impressed.
Web mapping is beginning to show signs of getting beyond the infantile and maturing from its pubescent phase and this example shows what can be achieved when you consider the entire user experience.
The maps are clean and well produced and there are plenty of them to support the inquiring mind, each accessible from a sensible tabbed box in the upper right. There's a permanent legend in the bottom right with not only a clear illustration of the chosen classification and shading schemes but a short description to assist interpretation. Nice to see data attribution and sources cited too. The title panel is sensible and although containing all the usual share and contact buttons is relatively unobtrusive. The graph in the bottom left is a masterstroke - it's linked to the map so we get a good scattergraph overview of the data distribution. Hovers provide the data summary and clicking a component in the graph orients the map appropriately. I really like the use of subtle graphical cues such as a slight animation to show an active element, or the emerging horizontal or vertical lines to anchor your eyes to the x or y axis. Likewise, hover controls on the map also deliver data summaries and the addition of a graphical yellow glow also gives focus. The ability to switch labels on and off easily also gives both unencumbered and contextual view of the map.
I also like the use of data re-apportionment into a consistent regularly tesselated grid which overcomes the problems of trying to use different geographies. It also makes moving between maps easier and supports visual comparisons more readily.
All that said, I'm going to get picky (because that's the purpose of the blog). I found myself frustrated by some of the cartographic choices.
Firstly, while diverging colour schemes tend to make a map look more interesting (more colours) it doesn't fit the data in a cognitive sense. Most of the datasets would benefit from a single hue progression or similar. Most of the variables are mapped with some arbitrary break defined where one colour morphs into another yet the importance of that critical middle value is never established. Is it important? The use of a diverging colour scheme suggests so but it is unlikely.
In fact, perusing through the maps shows an inconsistency in the graphical treatment. Most are diverging, some are single hue, others are multi-spectral (agh!).
Second, while the use of a regular grid is great the use of 3D on most of the maps is distracting. It's effectively a plan oblique representation of the hexagonal grid using a second variable to map population or employment density. Fine in principle and allows the map to remain planimetric (thus preserving scale across the map) but where you see large numbers of tall prisms it inevitably obscures a lot of detail behind. Prism maps have always suffered this limitation and I can understand that mapping the second variable gives us an important additional piece of information but it's questionable whether the cost of occlusion warrants it. The answer would be to include an ability to view the map from multiple orientations either through a rotate tool or just giving us, say, four of the cardinal compass directions. At least that way the map reader can see what's behind a block of prisms through map interaction.
Finally, the map works on multiple devices and some of the overlying boxes can be minimized - but not all. This does create a cramped feel on some devices and it would be nice for there to be more control over the position and visibility of these.
Like I said, I'm being picky but I'd like to see the cartography match the levels of the overall app, particularly in the use of colour.