Why Citymapper will be bigger than Uber XXXXL

The first time I saw Citymapper was at a modest stand at a Silicon Milk Roundabout. Since then it has become the much loved app that gets people from A to B in many cities around the world.

How can such a polished app and service be sustained? With the experimental bus service – and now the Black Bus project – the scope of their ambition has become clear and things have progressed faster than I’ve drafted this post. I’ll try to catch up now.

The gap

The key innovation of Uber is simply to use mobile Internet and GPS to match vehicles to journeys. However, in my limited experience, this doesn’t naturally extend to sharing which makes it difficult to take advantage of the efficiency scaling from filling a larger vehicle.

Busses are made for sharing but they don’t have the data to ensure they are in the right place at the right time – because there’s never been viable way to track real-time and fine-grained demand before.

As a result, you either hire the entire vehicle plus driver or share it with dozens, perhaps hundreds of people on a predetermined schedule. There is really nothing in between and this is the gap that Citymapper is targeting. It’s a massive chasm.

To make this happen, you first need to recruit a large population of sensors that are reporting travel currents in real time. This lets you construct a model of a city’s regular travel flows (Simcity) as well as honing your routing algorithms (ready for later).

From this, you have a unique insight into where demand is exceeding the supply of options, right now and with a large degree of predictability in the near future – people have to get home, after all!

Wheels on road

Software models in place, you need to put wheels on the road to fully make use of this insight.

And so you begin trialling actual bus services. It’s fiddly and different and you are only doing simple, fixed routes for now. But it’s a hugely valuable experiment where you find out what you don’t know about the practical logistics of registering app users as passengers on vehicles. You are also able to point at real numbers for time and money saved.

Partnering with an Uber competitor (Uber is not collaborative) lets you scale up the number of vehicles and passengers you are tracking - albeit on static routes still.

At some point you are ready for your next major step: on-demand minibus services. Effectively operating like Uber sharing, and needing a private hire licence, it will nonetheless feel like a magically serendipitous bus service. This is where I get a little speculative, for now.

The SmartBus

Citymapper knows your routine: where you live, work and play. It can make pretty good guess about where you go to the gym and what time you like to get home. This gives it singular insights and the potential to make the tedious business of getting around much easier.

If you’ve ever used a bus then you’ve probably run for a bus. The driver might have even waited a few seconds for you. It’s a good feeling when that happens. What if you got a notification from Citymapper asking when you were likely to leave the pub? Give it a time and it would swing a bus by a nearby Spot that’s easy to find and accessible to main routes.

Nothing beats trains and planes for moving a lot of people directly from one city to another. On a city road, a big bus is the most efficient. As you move further away from the bustle, the number of people making a similar journey to yours around this time shrinks and so must the vehicles. Travelling to the suburbs late at night may require an Uber. If you live in a village in the middle of nowhere then maybe you even need to own a car.

The actual problem

What are Citymapper optimising? Transit efficiency in aggregate. Less abstractly, they are targeting empty seats. Empty seats on off-peak busses. Empty seats in driver-only cars. Empty seats in parked cars lining residential streets. An empty seat is a sign a vehicle is taking up more space than it needs and is slowing everyone down in a congested city.

Ultimately, an empty vehicle is an anomaly. A vehicle should rarely be stationary – maybe only if it is charging or passengers are getting in/out.

A network becomes far more efficient as it becomes more connected with more options available for the algorithm to search.

Down the road

Ultimately, as with every transit technology shift, the dominant options reshape the city. Parking spaces will go the way of mews. Instead of home and work being clustered around stations, they will be able to spread to any easily reachable smart stops which rise from tracked demand. If there are passengers then transit options will appear. The adaptive routing algorithms will dynamically shape the route network literally overnight, rather like Internet routing.

New construction work that supports this routing will be able to assume that transit options will quickly and efficiently adapt to it. I imagine it’ll look like current small-world, fractal networks but with less street parking and more pickup points.

Self-driving cars – when they come – will be perfectly suited to this always-moving, anticipatory, coordinated model. Manual driving will be seen rather like horse-riding on the roads: not really practical and a bit of a liability.

The next problem is just working out where people need to be.