Google knows you took the bus: on the creepy accuracy of Google Maps Timeline

You are here. And here. And here. And... Image: Google Maps.

Knowledge is power, they used to say. Nowadays, they say “data is power”, and they’re not wrong. Unlike many of the modern, high-value tradable goods in our society like oil or gold, data is a limitless resource that we’re constantly creating more of day after day.

What the actors who own this data choose to do with it can often be a point of vast contention: should I be happy for Google to reliably know where I am, where I’ve been, and most frighteningly, where I’m going? It’s not up for dispute that the scope of these tools can be immense – but how much of that scope should we take for granted?

Google Maps is a tool full of wonderful surprises. It can plan a journey for you, tell you what deals to get at the supermarket, and give you updates at the bus stop. Some of the things Maps can do, it does without us even asking; Google knows when we pop to the shops, or when we stand by a bus stop.

This concept is called “geofencing”: cross-referencing geolocation data with the services at that location, and issuing notifications to a device on that basis. Google knows I’m in the supermarket because my location matches up with the area the supermarket is known to occupy, and through a complex series of phone masts and wifi access points, it knows I’m between the vegetable aisles. Okay, maybe things aren’t quite that specific, but the detail is stellar – and often, slightly concerning.

A simple flick through the timeline feature of Google Maps reveals that Google can plot day by day where you were, when you went home, and, maddeningly, how you took that journey – or at least, it can make an educated guess. By applying geofencing programming, Google can calculate when we are near a bus stop, and cross-reference that data with bus routes and other bus stops to determine with a reasonable degree of certainty when its users are taking the bus. Google doesn’t go as far as to try and guess which bus, but it could make an educated guess.

The same is true of train stations; pause in one, follow the expected route of the railway line, and travel through additional train stations, and Google will have no trouble in informing you after the fact that you have travelled by train. A reminder that you don’t need to have planned a journey on Maps for Google to surmise this: it is all calculated based on shifting geolocation data, and nothing more.

Walking, cycling and driving are harder for Google to calculate, because there are no geofenced points of entry for these modes of transport. It is therefore likely that, once bus, train and metro have been eliminated from the mix, Google simply inspects the time taken between harvested geolocation data to calculate the transport mode used. But without geofencing, it’s harder to determine the exact route taken by a user: because they’re not following a prescribed route, and because geolocation data is much easier to take while stationary, routes on timeline taken independent of public transport can end up looking… messy.

Google fails to surmise that some of this journey was taken by train and presumes I took an unorthodox drive through Kent in the early hours. Image: Google Maps/CityMetric.

The system isn’t perfect. For one, it can’t account for anomalies. I took a rail replacement bus service recently, and Timeline was dumbstruck by how I’d managed to get home. But the ever-increasing availability of data surrounding transport timetables means that the assumptions Google can make about our transport choices are only bound to get more accurate. That’s important, because its information that few organisations beyond Google are likely to have real access to.

If we take London as an example, we know that Transport for London (TfL) can use data on traffic flows, ticket barriers, and incomes for bus routes to determine how people use a service. In fact, TfL has even used its own wifi services to calculate route maps on the Tube. However, without undergoing intricate surveys, they will struggle to plot exactly how journeys are taken beyond the Tube Map, especially with regards to buses, disparately owned NR services, and so on.


Google has exactly the information to remedy this – and it’s integrated into Timeline, simply because people consented to having their location data collected. If your local borough council asked to do the same, and the only provision it could grant was that you might get a better bus service, many people would probably opt-out. Part of the reason why we accommodate the location-harvesting of Google is because we consider Maps such a vital service, and its domain – at least in terms of its rights to record our geolocation – is hardly contested. Even those of us who use Citimapper regularly tend to have Maps downloaded on our phone.

Google Maps is in a unique position to mark the differences between journeys that are entirely spontaneous and journeys that are pre-planned, because it is measuring both. That information could be highly useful in designing timetables and shaping user-friendly services.

Moreover, as geolocation data grows more precise, it will be able to help us pin down the flows of pedestrians and cyclists in our cities. While it’s possible to gather this data in the public domain without geolocation, it’s economically prohibitive to do so in less densely populated areas. This data would help prioritise cycle-friendly and pedestrian friendly developments on the understanding of where demand is greatest.

This sort of data inevitably carries such a high risk factor, however – not only as far as personal privacy is concerned, but also surrounding efficacy. We presume that if we know every individual's travel patterns, we can design perfect travel services – but patterns change all the time. An algorithm can never incorporate the latest change before it is registered by the system. While data like that collected by Google Timeline could be put to better use by transport authorities, it shouldn't be abused, nor serve as a panacea for good design.

Worst of all, it’s hardly clear that this data is up for public consumption. The furore over data protection means it would be considered deeply unethical for Google to hand this location data over to anyone, let alone a local government body like TfL. It may be moot point; Google itself claims that Timeline is for our own amusement and little more.

But maybe we’d get better services if it wasn’t; after all, geolocation isn’t slowing down anytime soon.

 
 
 
 

What we're reading: Understanding how the coronavirus spreads in public spaces

Risk assessment: It’s a holiday weekend in the US and UK, and where the weather is nice, people will surely want to go out. Vox has a handy chart for understanding the risks of coronavirus in different settings.

Covid-proofing: Social distancing has proven to be an effective way to slow the coronavirus, but it’s an emergency method that can’t stay in place forever. In order to get the economy going again, offices, restaurants and entertainment venues will need a dramatic overhaul. The Atlantic shares ideas to make that happen.

Vacation ghost town: With no indication of when people can safely travel again, resort towns are bracing for a summer unlike any other. CityLab reports that this weekend is the start of a critical period for vacation hotspots, but residents and businesses there expect tough times ahead.