23 Nov The science of location
One of my favourite coffee table books is Powers of Ten. It shows our universe from the very smallest to the very largest (a similar video is here), and if ever you wanted grounding in what’s going on around you vs. everything else it’s just a 3-minute watch and suitably humbling.
It also reminds me just how far we’ve come in understanding and measuring everything that is around us (and just how much we still have to do). Last time, I touched on the history of location and how we’ve come to have billions of people with billions of devices all capable of knowing exactly where they are. Every smartphone now has access to GPS, Wi-Fi and 3G/4G combined with unprecedented levels of processing power, connectivity and cloud based services to not only know a geographical position but the actual place of 2 billion people. Wow.
Blue Circle vs. Blue Dot – and why accuracy matters
So why, when you exit the tube for that meeting, does Google Maps send you off one way down a street, only for that blue dot to race to the other side of the block?
To answer this, we need to look at how we get location, what accuracy really means and why it’s important to different uses of location tech. Most phones now use GPS as the primary method for obtaining an accurate location. As I touched on last time, GPS or “Global Positioning System” is the US military system to provide accuracy down to a metre or less – or does it?
GPS relies on the receiving device seeing at least 2 orbital satellites in order to get an accurate 2D fix – your latitude or longitude. For 3D (i.e. your altitude, often translated to the height above the current terrain) it needs 4 or more. Finding these satellites and getting an accurate signal from them can take time, particularly if you’re starting from scratch and don’t know where to look. Most people will have experienced this with car sat navs of last decade, when they’re first switched on (and have moved from their last location) they can take a few minutes to look up and find the right satellites for that particular area – 12.5 minutes for a “cold start” as the data rate of GPS is some 50 bits/s.
As most people can’t wait that long, especially when emerging from a tube station and late for a meeting, smartphones have a few tricks up their sleeve to speed this us. Firstly, they maintain their approximately position using cell tower data and Wi-Fi look-up. This means they know roughly where to start looking for GPS signals – they get a head start and the “time to fix” is reduced. Secondly, and most importantly, they use Assisted GPS, which in its simplest form passes off the GPS data caching and processing to a nearby cell tower.
However, A-GPS, cell tower triangulation and a few other tricks don’t always work perfectly. This gives rise to the “blue circle”, essentially the ‘area of uncertainty’ (or level of accuracy). We’ve all seen it – quite often when in a large building or underground station. As the phone picks up more signals, it often shrinks down, hopefully to the blue dot, and you can be on your way to that meeting.
Or so you thought – suddenly that blue dot charges off down the street, or like some Marvel superhero dashes from one place to another. This is quite often when the GPS signals are being reflected or blocked by buildings, given rise to “canyoning” or a false position. A good example is shown below.
The image below also shows how different GPS systems perform against the “ground truth”, i.e. the actual path of the consumer’s device. GPS is shown as red dots, GPS augmented with newer satellite positioning systems such as GLONASS and GZSS are the other two colours.
Location Sciences sees both the accuracy and inaccuracy in our location data – and we embrace these differences. Because we see the precise path of a device, including altitude, speed, bearing and combine this with the horizontal latitude and longitude, we can use known inaccuracies, correct them where necessary and then learn from them. If one or two points in a path deviate, we can learn where to correct, based on previous behaviour, and because we constantly collect this information, anonymised and securely, direct from the device ourselves, we can verify and authenticate it to a very high degree of accuracy.
Combine this with the essential requirement to see the data 24×7 (so not just via periodic ad requests, social feeds or server logs), we can build a “location graph” for a consumers device. This location graph of a device (and of many millions of devices) can then be used with confidence to measure consumers interaction with places – e.g. did these people walk past a store or actually visit it? Did they travel on a bus down the Strand or walk? Location Sciences can answer these questions in detail, but also simultaneous at scale across our millions of consumers and billions of data points.
So, accurate location data, preferably 1st party location data derived directly from the device using specialist SDKs, will get you a long way, but what’s really important is the science that combines with this source data. An example of this is Location Sciences store visits analysis, which uses the path of many consumer devices and their reported accuracy to determine the probability of a store visit, rather than just whether a number of events has been recorded intersected with a store’s lat/long.
Until the next instalment…
Chief Strategy Officer