Citizen-as-a-Sensor

What does that mean?

I came across the term yesterday for the first time, and wanted to know more. IBM were showcasing their product offerings that use it, and I stopped for a chat.

IBM had been tasked with identifying citizen movement trends in Istanbul to assist in that city’s future planning of roads and public transport. The city officials needed to know how, where and when the masses move. In smaller areas this can be done via aerial tracking and sensors (for vehicular traffic) and manually or via turnstile tracking for public transport. However, having to monitor millions of people travelling in every direction IBM took novel approach using mobile carrier location data. Partnering with Vodafone, IBM got access to the mobile carrier’s Home Location Register (HLR) data and carried out Big Data analysis to establish the people movement trends they needed.

In addition to the obvious benefits for planning public transport and roads, people movement data can be handy for retail strategies (branch locations and open hours), traffic planning, insurance, public safety (where and when to provide security), road works planning, etc. It could also provide interesting insights in to crowd behaviour following irregular incidents/event s (weather events, public disturbances, major sports events, etc.)

This is ‘treasure from trash’ solution.... or rather more akin to producing energy out of corn husks or renting out your back yard for your neighbour's cows to graze J. Normally, the masses of HLR data would be give limited value, and only just for the mobile carrier. The mobile carriers would be interested with the numbers within an area at various times, but would be less concerned with  the movements of individual users. Opening up this data to government and others creates a potential new revenue stream for a Telco, and broad benefits for the public (by way of more targeted services).

So what are the limitations?

The most obvious is that not everyone has a mobile phone, hence students and lower income people may be under-represented in the numbers. Secondly, unless you get all the mobile carriers you only get a sample. Another is that you need to perform some hefty Big Data analysis, which is a specialized area and could be costly. Hence, various assumptions will need to be made and possibly some tweaking of the data.

Overall, when speed (translating to cost savings and quicker solutions) is more important than accuracy, this is a very good method.