Bogota: The Price of Service Complexity

(Leer en español.)

If you look for a map of Bogotá’s TransMilenio network, you’ll find something like this:

Diagram of TransMilenio from the City of Bogotá website, published by Jose Luis Martinez.

Note that north is to the left on most standard Bogotá maps.

This is not, of course, a map of the whole network. It barely even begins to explain what the buses do. In fact, what the buses do — even the biggest buses providing high-capacity transit — is so complicated that it may impossible to map clearly.

This map invokes a familiar stylistic language from metro maps worldwide: A solid line of a certain color means direct service between all the dots on that line. Where two lines cross at the same station dot, that means you can connect between those services there. The style is so pervasive and effective that some cities without much rail service draw their bus services this way. (Here’s the bus map in Wellington, New Zealand, for example). Just down the hill from Bogotá, the Medellín Metro system uses this familiar style to show how their rail, bus rapid transit, and gondolas all interact.

But Bogotá’s map doesn’t work that way. No logical metro network would have lines like B, A, and H as they appear on this map, meeting end to end. In the usual metro map language this would mean you have to transfer twice to keep going in the same direction, which would be silly. But that’s not what it means here.

This is a map of branded corridors of buses, a concept that takes some explaining. The green line B on the map means that along this corridor, a bunch of buses whose line numbers start with B all operate. But those buses are coming from points all across the network, bearing their B number.

Here’s the explanatory sign of a single station, Virrey on the B segment.  (Click any photo to enlarge and sharpen.)

This sign lists every bus route that stops at this station, with a list of the other stations that each route serves. The colors match the map. We’re on the green “B” segment so sure enough, there a lot of green B bus routes. But those buses are nowhere near the majority of all the buses that stop here. In fact, if you’re traveling between two B stations, the next bus that does that may not be a B bus at all. What’s more, the next B bus may not take you to the B station you want, because of their complicated skip-stop patterns. So you really have to check that the bus you get on is stopping exactly where you want to stop.

Why so many different patterns? This is the compounded effect of two kinds of complexity:

• They are trying to run direct service from this corridor to every other corridor in the network, so people don’t have to transfer.
• Not all buses stop at all stations, so stopping patterns generate separate routes even among buses that are all following the same path.

Notice that both of these effects are multiplicative rather than additive. If you wanted direct service between any two or the 12 corridors, that in itself would generate 72 possible routes (12 squared is 144 one-way pairings, which is 72 two-way routes). Then when you start varying stopping patterns you continue to multiply routes. Remember, more routes always means less frequency on each one. It means more time standing in a crowded station before you can start moving toward your destination.

In the lower left of the sign above, there are also six bus routes numbered 1-8, with headings in black. These were added later. These buses are distinctive because they stop at every station, so they provide a simpler service pattern. (It would have been nice to have a map of at least these eight routes, which are clearly simple enough that such a map would be legible.) But these routes are a small part of the network. Most of the great masses of red buses that you see in the TransMilenio corridors are doing one of the more complicated patterns.

Most of these patterns run with a frequency of 5-10 minutes, but of course there are disruptions, especially in routes that run partly outside the infrastructure, so I often waited longer. Ten minutes is a reasonable wait in the transit-starved United States, but on TransMilenio, with many buses per minute flowing past each station, and given the discomfort of most of the stations, it feels like a very long time, especially if the complexity is also heightening your anxiety.

So let’s review all the ways that this complexity affects the rider and the experience of TransMilenio as a whole:

• It requires longer waits. In simpler systems, there are fewer service patterns but they run more frequently, so people are more likely to get on the next bus that comes and make progress sooner on their journey, even if that bus doesn’t take them all the way to their destination.
• This means that stations are more crowded, as more people wait for a particular bus instead of getting on the next one.
• It makes buses more crowded because they are more unevenly loaded, for the same reason. This is likely the cause of the difference between TransMilenio’s “theoretical maximum” capacity (56,000 people per peak hour per direction) and the highest actually observed volume (43,000).
• It makes the station wayfinding challenge very difficult, because people have to be navigated to a particular platform within the station, and the huge number of routes means that the user must wade through a lot of information (see sign above).
• Most importantly, it trains people to either (a) rely on an app to navigate them or (b) use only services that they know. Both options also tend to steer people away from other services that might be more useful to them for the trip they’re making.

The complexity is staggering enough if you travel within the TransMilenio busways, but once you’re outside of them, it’s truly overwhelming. Here’s a bus stop sign in a very dense inner city neighborhood.

That’s a lot of flavors of buses, all going down the same street in the same general direction. You already know that route numbers beginning with A, B, and C mean that those buses will eventually end up running on those corridors of the busway, though it’s anyone’s guess whether that segment is in the future or the past relative to this sign. And there are so many other routes on top of each other. This is what the network looks like at street level, stripped of the relative clarity of the big busway stations. There is information here — Bogotá’s numbered grid system means that a practiced rider can decode much of it — but the cognitive load is immense, and for a newcomer it is simply a wall of text.

To TransMilenio’s credit, they’ve done all they can on a static sign to explain these bus routes with text. Bogotá is blessed with a grid system of numbered streets, called carreras when they run north-south and calles when they run east west. This allows routes to be described very briefly, because locals learn the code: K for carrera and C for calle. (A for Avenida just means that the calle or carrera is a major street.) So there is a great deal of information here, which I quickly learned to make the most of. But it’s still a lot of complexity.

Later, I attempted a trip from K11 & C92 to K69 & C24. The journey required a transfer, which was fine with me, but Google told me to transfer at a point where there was only one bus route continuing to my destination. As I waited 20 minute for that bus, countless other buses flowed past me, each one maybe taking me partway to my destination, and to a place where there might be more buses to where I was going. But there was no way to make sense of this information.

As I understand, the local app TransMiApp will give you directions based on where the buses actually are, so that for example if my direct bus were late it would serve up a faster alternative involving a transfer. But the right answer would continue to change by the minute during my journey as various buses ran on time or not. This would have helped, but not nearly as much as having a simpler system of fewer routes, all running much more frequently. There are enough buses in Bogotá so that on almost every street, the next bus could always be coming within a minute or two, ready to take people on the first stage of their journey. But that would require a different approach to design, one that is less fearful of transfers.

Let’s acknowledge that Bogotá can be critiqued on this score only because:

• They have unified planning control of the whole network, so they theoretically have the power to make things simpler.
• They have at least attempted to explain the service, which makes it possible to see what’s wrong.

In many middle-wealth countries, and certainly in poorer ones, you won’t see any effort to make the service legible, so the complexity is overwhelming that it’s hard to even talk about. In many countries, too, private operating companies still control service planning and do not feel motivated to participate in creating a logical total network. If I seem to be picking on Bogotá here, it’s only because unlike so many of its peers, its public transport authority is at least trying to present a network, and it has enough planning authority that it can be held responsible for that network’s design.

Still, if I had been working on this network, I would have questioned how much of this complexity was truly necessary. I would especially have asked:

• How much of this complexity is needed only in the peak? One of the most common mistakes in transit planning is to design the all-day network around the rush-hour pattern of demand. Instead, the all-day network should always be designed around the all-day pattern of demand, with rush-hour patterns added on top only as demand warrants. That can help you create a more legible all-the-time network — maybe even one simple enough to draw a diagram of — even if rush hour service is adding a lot of complexity. In our firm‘s maps, we draw rush hour services very faintly, if at all, so that they don’t distract from the simplicity of the all-the-time pattern.
• Where in the network is the infrastructure for changing buses so good that there is no reason to insist on providing a single-seat ride? There are already several convenient interchange stations in the network. Asking people to change buses is always the key to a simpler and more frequent network.
• Can we make simplifying assumptions about stopping patterns to reduce the number of service patterns? This massively complex system has millions of details that can be fine-tuned to demand, but what is the cost/benefit ratio of keeping track of all those details. Would a simpler system that’s easier to understand produce benefits that outweigh that?

There’s one easy response to these questions:  “People hate to transfer!”  In Bogotá the decision to avoid transfers is based on surveys in which people are asked if they want to transfer.  But these surveys asked a hypothetical question without explaining consequences.  It’s always important to ask:  “Which matters more to you?  Avoiding a transfer or getting to your destination sooner?”  Because that is often the real choice.