Makeover of SPC Charts

Statistical process control (or SPC) charts are designed to show variation in predictable data. Variation outside of a set of control limits is considered an anomaly, or signal. They are used by West Midlands Police to track crime rates and others key performance indicators across the force. One of their main difficulties with existing SPC charts is the ability to identify geographical relationships within the data. For example, whether two policing neighbourhoods that are spatially close exhibit similar behaviour.

This project involved a redesign such that multiple SPC charts, faceted geographically, could be viewed concurrently. The first step in this process was to develop an SPC chart and implement the signal and process-break detection, which I did in d3, and you can play with below. Click on a data point to disable it from the analysis, and click in the white space to insert a process break.

Chart 1 – An SPC chart of crime in the West Midlands, UK, aggregated by month. The solid green line is the mean, and the dashed lines either side are the control limits at 1.5, 2 and 3 standard deviations from the mean.

As we entered a process of redesign, we wanted a method for regularly updating analysts at WMP with our ideas. As a solution to this, I developed what we have now coined Dynamic Design Documents, which are a series of interactive web pages that describe the design reasoning process and allow data analysts and West Midlands police to play with our redesigns in their own environments. You can see the dynamic design documents here. Our work on dynamic design documents won best the pest poster award at the 2017 IEEE VIS conference (here is a link to our video preview that will be shown during the conference).

The design documents will take you through our design rationale, but you can also see one of our designs below. The gridded map represents the 174 policing neighbourhoods in the West Midlands, arranged such that they are similar in position to their original geography. Within in each of the neighbourhoods, analysts are able to see (i) whether a signal has been triggered, (ii) its signal history, (iii) its process history, and (iv) the current trend of the data.

Chart 2 – A grid map representation of 174 SPC charts – one for each police neighbourhood in the West Midlands. Click on a square to explore its corresponding SPC chart.

Furthermore, I have taken the grid design and created an interactive application where analysts can upload and explore their own data. This is now deployed at West Midlands Police and can be used by analysts to help in their day-to-day activities.