AGI & CDRC Geo+Data 2018

Last night (2 Oct 2018) I attended the first in a series of events held by the Association for Geographic Information (AGI) and Consumer Data Research Centre (CDRC).

Previous events by CDRC were typically very high quality, and this was no different. Hosted at UCL London, the evening session took the form of a brief introduction, followed by two talks, and socialising. Ideal for London-based students and interested parties, although worthwhile enough for me to travel up from Portsmouth as well.

The first session was presented by Alastair McMahon, of Telefónica (O2). Alastair shared some of the many datasets that are becoming available through the use of mobile phones, from a network operator perspective.

Your mobile operator can potentially gather an enormous amount of information about you, your interests and location. O2 have demographic information from their users. They know which websites you visit on mobile, and can trace with reasonable accuracy the movements of your phone over a wide area.

Compare this to municipal geolocation networks (such as wifi) which can only monitor movement within a relatively constrained area (the town) – O2 clearly have UK-wide coverage (and likely further afield as well…)

The packages of data, all anonymised, include dwell time, origin-destination, demographics and movement. This is an enormous trove of information for retailers and place management specialists.

So why aren’t we using it already? As one audience member commented, these things are still prohibitively expensive. I suspect there will be a lot of movement here in the near future as mobile phone companies improve their service lines and packaging of data, so look forward to more of an ‘off-the-shelf’ data provision from mobile companies. Right now, it’s an enormous set of data (4-5 billion points per day) to manage and commercialise.

Meanwhile, there are plenty of research opportunities. The image above is an infrastructure connectivity map of England and Wales. It shows – in wonderful detail – how readily connected the major urban areas, such as London, Birmingham and Liverpool/Manchester/Leeds are. The deep blues of mid-Wales and patches around the country show areas where connectivity is poor.

Session two was by Bala Soundararaj, a PhD student at UCL. Bala has worked extensively on the SmartStreetSensor project with LDC to calculate footfall from mobile phone data – a subject close to my heart as well.

In this session, Bala described the techniques and challenges involved in collecting anonymised pedestrian data from a series of phone sensors. This system, like most phone sensing networks, uses periodic signals sent from mobile phones known as probe requests. Every device with wifi capability sends out these signals.

The challenge is to identify one phone from another. A few years ago this was not an issue, as each device had a unique code known as a MAC address which it sent out with every request. However, last year Apple and Google both made significant changes to their phones to prevent this kind of de-duplication.

Bala then described his workaround, to use a combination of signal strength & what is known as sequence numbering, to identify one user from another. This retained a degree of confidence in their results and enabled the sensors to keep providing meaningful figures even after the changes by the phone manufacturers.

As phone sensing is being used increasingly in public spaces, it was interesting to gain some insight into the methods used (the project is national, with over 1500 sensors in place around the UK). Furthermore, it was useful & reassuring to compare with my own work. All in, a very interesting presentation.

The Geo+Data event was one of six planned through 2018 and 2019. The next is on 4th December. If you are working in retail geographical analysis or a student in this area, I do recommend it.

Sven Latham

Sven Latham

Stares at spreadsheets for a living