This clustered force graph (somewhat inspired by this) shows relative popularity of toys from toys data obtained at one toy library using SETLS. The 'popularity' is based on the number of times the toy has been loaned, though this may not be a true reflection of its actual popularity. Other data to measure popularity we are capable of obtaining is by percent of time out (versus time toy has been owened by the library) which is a far more robust measure. However, the need to prototype this graph, we are only basing this on the number of total loans each toy has had.
The actual data file contains toy popularity information for over 3000 toys, but a good bubble chart this does not make! The size of the circles must be relative to each other, so once you start trying to determine the size of a circle representing over 2000 loans versus something that has been loaned once, means that we can't really use any visual way to show some of the data.
Based on experience exploring the data, about 130 circles plots well. The data is then binned into histograms using D3's own d3.bin.histogram function where we decide on a maximum number of bins. I've let this be 20 bins - then count backwards from the highest binned number till 130 datapoints is reached - the relevant toys are selected and plotted.
Each toy has three possible descriptors (manufacturer, age range, and toy category --defined by the library). Each toy library has vastly different categories, manufacturers, age-ranges, gaps in their data as well as size of their data - this seems to be a fairly robust solution across 53 toy libraries for now!