An example of guidebook routing
An example of guidebook routing

Lots of interesting things have been announced at Google I/O this week, including a major update to Google Maps, a Google product that’s familiar to almost everyone, and used by many on a daily basis. Most of the news coverage has revolved around visual, social or privacy aspects of the Maps experience, but I want to talk about a major upgrade to the transit functionality of Maps.*

With the new version of Google Maps, when you ask for directions between two points, rather than getting an itinerary that minimizes travel time for a handful of particular departure or arrival times (as you do today), you’ll be offered an itinerary that gets you between those points, as frequently as possible, for as much of the day as possible.

To put it in transit nerd terms, Maps will evaluate all the possible ways to get between two points to figure out the effective all-day frequency and span of service (accounting for connections between services of different frequency), and show you itineraries which prioritize those qualities over a naive minimization of scheduled travel time. It will still be possible to look at departures or arrivals at specific times, but the general guidebook itineraries will be the first thing users see.

The screenshot at the top, taken from the public preview, shows an example of this. To travel on transit from the PacMed building to downtown Fremont, take bus 36 and then transfer to 26, 28 or 40 in Pioneer Square. This itinerary works at least every 15 minutes from 6AM to 11PM, every day; within those time periods, it’s a general solution to the problem of getting between those two points. An alternative route, using the 5, 16 or 26X to get off at 38th & Bridge Way and walk down the hill is also available.

In both cases, note that even though a single route determines the baseline daytime frequency for the connection, Maps notices that other routes also serve an identical pair of stops origin destination stops, so if one of them comes first, you should take it.

After the jump, another example.

seattle prep vet

There are three ways you could get between these two places: walk to Eastlake and take the 70-series buses, or the 66, or wait for the 25, which makes this trip almost door to door. The 70-series runs every 15 minutes all day and night, the 66 is every 30 minutes, and the 25 is every 60 minutes. Nobody capable of walking is going to bother waiting up to an hour for the 25, and the results reflect that, directing you towards the most frequent service, rather than the nominally fastest service, which would be the 25 at 19 minutes. If you really do want to see the door-to-door trips, you can click the “More options and times” links and find them.

This new transit routing system more closely reflects the way the vast majority of people actually use transit, prioritizing frequency and span of service on generally-useful routes over infrequent one-seat rides that just happen to provide doorstep service. Its works best, of course, in transit systems full of truly high-frequency, reliable all-day routes, but to the extent that we have such service here, the same ideas apply. Frequency is freedom, and the most commonly-used private sector mapping tool increasingly reflects that. There are still some kinks to iron out of the new routing engine**, but this is a huge stop forward to more comprehensible transit.

The new Maps is being rolled out gradually; sign up here to get on the waiting list. I think you’ll love the new maps.

* Disclaimer: I work at Google, but in a completely unrelated area, and all this information is based on the public preview of the new Maps, not inside knowledge. Please don’t ask me questions about unreleased Google products or features, I can’t answer them.

** For example, the 7 is prioritized over Link for Rainier Valley trips because Link’s all-day frequency is considered to be 20 minutes, due to the post-9pm single-tracking that’s currently happening for repair work. The 7 runs every 15 minutes until much later, and is thus considered a superior service by the routing engine, even though during most of the day it’s is actually an inferior service for most trips.

34 Replies to “Google Maps Introduces New, Smarter Transit Routing”

  1. Nice. This is a very nice feature, but I think Metro has had something similar for many years (e. g. you can specify how much walking you are willing to do, etc.). I know there are differences, but I think the best part about this is that you see a Google Map in the background. As I mentioned in the previous post, I wish more people would use Google Maps as their default background map (instead of writing a PDF).

    On an unrelated note, I wish Google would add more map layers, including a USGS (or other country) topographic layer. The Terrain layer is not very useful when it comes to hiking. This is why folks make sites like ACME ( or GMap 4 ( GMap 4 is made by a local guy (and is the best I know of for this sort of thing).

    1. Metro’s trip planner suffers from the same weakness as all conventional trip planners: it asks you when your trip will take place, and offers you a micro-optimized route for that trip, assuming that every bus always runs exactly on schedule.

      Google’s huge innovation here is giving transit directions more like highway directions: they give you the best route to take *in general*, based on frequency and span of service, rather than individual trips. I’ve never seen any other transit planning site which provides this, even though it’s clearly far more useful than the alternative.

      The fact that you can select preferences, like minimizing walking or preferring trains, is just icing on the cake.

      1. Actually, I think Google has switched to suggesting the best highway route to use in current traffic conditions, not in general. This is a good idea, but I’m not sure how to turn it off (maybe you have more options with the new Google Map). It is a good idea because you don’t want to send people to the express lanes when they are going the wrong way.

        Also, it is pretty easy to go from a particular trip to a general one using the Metro app. I do that all the time. You get to the same place, it’s just that you have to click a bit more. The Metro site shows various route options and allows you to put in maximum walking distance, fewest transfers, etc. None of these systems are perfect because there are variable uncontrollable factors (such as traffic). But either system gives you a nice range of options. The same is true for driving as well. Google route finding is great, but there are numerous examples of it steering people wrong — nothing terrible, it just sends people on routes that are slower than what a knowledgeable person would use.

        The big selling point (from a usability standpoint) is that this is using and leveraging Google Maps. This means that you can just click on the map as your starting spot or destination (no fiddling with picking a spot from a list after you initially put in a street name or other landmark). The Metro UI is very clunky. As I said, you can get the same information, but you have to work harder to get it. My biggest concern is whether it will have a full range of options. As good as Google Maps is, it often lacks options I would like to leverage. My guess is that they prefer a simpler interface over a huge number of options. The new Google Maps may have more options while still keeping a pretty clean look.

      2. Also, it is pretty easy to go from a particular trip to a general one using the Metro app.

        I think you have a different definition of “easy” than I do.

        Here’s a sample Metro itinerary. Nowhere does it tell you the frequency of these routes. Nowhere does it tell you whether there are any alternate options. Nowhere does it tell you the span, so that you could know whether this trip actually exists for more than 5 minutes a day.

        Here’s another one, from 200 Broadway to 2420 Boyer. Again, no mention of the fact that the 25 only runs once every hour, and that catching the next 43 will actually make your trip about 60 minutes longer.

        In contrast, here’s the route that Google gives for the same trip. Not only is it a better route for most people, but it immediately tells you what you want to know: what’s the frequency, what’s the span, what’s the expected/reliable trip length, and when is the next bus?

        You can try to reverse-engineer that information from the tiny tidbits that the Metro trip planner gives you, but there’s a good chance it’s not possible, since the Trip Planner route will be micro-optimized for the specific time of day that you happened to request your trip to start. But even if you could, Google gives you that information *immediately*, which is leagues better than having to figure it out yourself.

  2. Major props to the Google Maps team for their excellent work. This is a huge win for transit legibility.

  3. Very cool idea. This is exactly what people need, much more than specific Trip Planner itineraries that are occasionally wrong and often suboptimal.

    Rail bias should be built in… especially given the likely audience.

    1. Wouldn’t a rail bias just fall out of the algorithm if rail provides a real advantage in terms of frequency, span of service or transit time? That’s as it should be.

      I notice that the Eastlake -> SLU trip example didn’t suggest a transfer to the SLUT.

      1. In a case like the SLUT, where there is no travel time or frequency advantage, I still think there should be some level of rail bias built in. Unfortunately the 40 is so far superior to the SLUT as it operates today that even built-in rail bias probably wouldn’t prefer the SLUT over the 40. (And the 40 will just get even better when it goes two-way on Westlake.)

      2. The problem is that the most obvious benefit of rail is in *schedule reliability*. And Google Maps is not so great at factoring that in, unless they manage to put an (“in traffic”) modifier on the bus routes.

      3. Schedule reliability comes from unobstructed ROW and or the ability to reroute, not vehicle technology.
        Metro North collision messed up the schedule there for several days, Link hits a car and schedule reliability is out the window… Bus bridges for blocked trains take time and added expense to put in place but rerouting the next bus coming along on a line to avoid an accident is a relatively simple, quick and reliable fix. It may add time to a trip, but headways can be maintained and downstream impacts minimized with a bit of skilled management. Rail built with sufficient crossover opportunities may be able to do some of that if only one track is blocked, but operating single-track is going to demand higher headway.
        The RR I used to work for regularly had 2-3 hr outages in the commute period because of someone who intentionally or stupidly got themselves run over. It took 2 hours to deploy a bus bridge so usually did no good at all. Buses suffer from accidents and traffic variation too, I’m not saying either is better, It’s not the vehicle, it’s where it operates.
        Is there a bias to service operating in exclusive ROW? That would make sense. So would a bias for level boarding, off board payment, signal priority, multiple entry/exit doors, consistent stop pattern (always stops at every station), etc…

  4. This will save me so much time searching. It can literally take me 20 minutes to figure out a new route using up to three sources, and this will do that optimization for me.

  5. I’d also point out that we’ve added a new Schedule Explorer for more easily comparing detailed transit itineraries, including a nice visual comparison of different trips (their durations, number of transfer, routes, etc). You can see a demo of it in action in the following YouTube video:

    1. When I saw that last night after I got my invite to the preview, I just stared at it in disbelief. Really cool feature!

  6. This is a giant leap forward for transit legibility, and for encouraging the use and expansion of frequent networks, especially in cities [cough, cough, this one] where frequent networks barely exist and one-seat bias is rampant.

    Bruce (or Brian), how is the highlighted result’s trip time being derived? Is it the mean/median journey, including the transfer penalty? Is it the maximum scheduled trip-time within the noted 6am-11pm period, including transfer? (I’m presuming wait time for the first leg is omitted.) And is it based on the “single route [that] determines the baseline daytime frequency for the connection”, or on all of the available routes?

    Hardly Google’s fault, but it’s still awful to see such a simple trip, with so many frequent-ish routes comprising it, expected to take 3/4 of an hour to complete. Especially when clicking the little car next to the transit symbol will drop that number to 13.

    1. One problem with Google Maps in this city is that its driving times are ludicrously optimistic. In Seattle, it appears to assume that you drive the speed limit on each street in your directions without ever having to stop at a light. So the difference is further exaggerated. (I have not found that to be the case in East Coast cities — there, I can typically beat the forecast time.)

      In the real world, there is no way you could complete that trip in 13 minutes in a car any time other than very late at night, and even then it would be iffy.

      1. Don’t forget that I was a Metro driver who ran on time 90% of the time…

        If you’re getting from PacMed to Fremont in 13 minutes during the day, you have a flying car.

      2. No, I have a car2go and I’m paying by the minute.

        We’re not talking about rush hour here. But it’s totally feasible at noon, either using Google’s clever loop to Dearborn or by taking Boren and zig-zagging down to the Cherry Street highway ramp.

        The key, really, would be making the lights as aggressively as possible in the four blocks you’re forced to be on Mercer.

      3. While I have sometimes felt the temptation, I really try not to drive more aggressively on a Car2Go than I would on any other vehicle. At best speeding and fudging yellow lights can save you a dollar or so. The fine for getting a traffic ticket and the emotional trauma of running someone over are worth far more than that.

      4. That was mostly a joke; I don’t actually drive any more aggressively in a car2go.

        I drive equally assertively in any car.

        What Seattleites never seem to get is that driving offensively — with an absolute awareness, applied to take advantage of cycle timing and available space — is both far more efficient and MUCH SAFER than driving on oblivious Pavlovian red/green autopilot the way Seattleites tend to.

        I’m convinced that if Seattle drove more like Boston, we’d eliminate half of our rush-hour gridlock. And guess what? You’re less likely to be injured as a driver or a pedestrian in Boston than anywhere else in the country!

        So, yes, I’m very proud to drive like a Bostonian. (By contrast, I’m a veritable wuss when driving on suburban highways, since I’ve so rarely done it.)

      5. Last time I was back in Boston I fell back into all the old habits. One of the ones that makes things move a bit faster (but would completely freak out the locals here) is the practice of the folks waiting to turn left jamming on the gas the second the light turns green (thereby cutting off the folks proceeding straight across the intersection). I actually did that on my driver’s test and the Commonwealth of Massachusetts still gave me a license.

      6. The “Pittsburgh left” fills me with rage… because I actually go when the light turns green, and the people who do it get in my way. I always give them a good long honk.

      7. Being a “driver” completely schooled in the NW tradition I couldn’t even imagine driving in NY or Bean Town. Nor Chicago where I did a short stint in school and I have to admit it was great to not to “have” to. I went to school with a friend that grew up on Long Island and did his undergrad (with a car) in Boston. He confirmed that back east; particularly in Boston, people jump the green. He, OTOH was green with fear when coming out to visit and being in a car where we routinely “pushed” the yellow. He was sure we would all die. In general I’d say that back east there is a much higher expectation that “the other guy” is paying attention. Out west… walk in front of a buffalo and it’s your own damn fault. Unfortunately that extends to bikes and pedestrians. Your SUV is the equivalent of a six shooter hanging from your belt.

      8. I see people “jump” the green that way in Seattle. There are certain intersections in particular where this happens… northbound Phinney at 65th is the first that comes to mind… so maybe it’s always the same guy. Maybe that’s d.p.’s route.

      9. No, I definitely never “jump the green” in that way, and no self-respecting Masshole would do something so dangerous in a situation where the other driver’s behavior can’t be estimated and eye contact can’t be made. As Bernie says, in Seattle you always have to assume the other driver is responding to Pavlovian cues rather than paying attention to their actual circumstances.

        I do still run yellows, especially on protected left turns (there’s a decent delay before the opposite straight gets a green).

        I definitely will pull way out into the intersection — not doing so blocks everyone behind you, especially buses, and is the quintessential indicator of Seattle’s lack of communally-beneficial adaptive action — and I’ll sneak through the moment there’s a break in oncoming traffic. (This isn’t rocket science, but it is physics!)

        But my most Seattle-anathema habits are the ones that involve weaving around slow-moving or stopped traffic, on-the-fly route re-optimizations that may involve sudden u-turns, and the classic “Oh? That’s not a lane? Well it is now!”

        I don’t drive all that much, but it should still be noted that I haven’t had an accident since I was 17, and have never done damage to another entity beyond a too-tight-parking-spot bumper bump.

      10. Traveling across 520 every day on the 542, I have come to appreciate Metro bus drivers’ abilities to always know which lane to be in and to be aggressive enough when making lane changes to get over where they need to be.

        And the fact that as a rider, I don’t need to feel stressed out, or guilty for cutting in front of a line of cars, makes it even better.

      11. I’m with you, ASDF.

        I’d always rather not be driving, so I greatly appreciate the Metro drivers who drive like I would: making strategic choices like you describe, and thereby doing their part to make the system run a little more effectively and a little less painfully deficient when compared with driving yourself.

        The inverse of this, however, is that I have no qualms about saying that those who drive like they don’t give a crap about passengers’ time do not deserve to keep their jobs. Drivers on the 44 eastbound, some of whom sit in the right lane for one… two… three light cycles when the left lane has been moving smoothly, are essentially holding passengers captive. Any driver who does this should be fired outright.

      12. It sounds like Metro drivers do the Chicago merge. In my experience on the east coast and in CA people don’t seem to know what the turn signal is for, they swerve back and forth whenever they feel like it. While most midwesterners are fairly nice drivers and signal to ask permission for the merge, in Chicago the signal is a friendly warning that I’m changing lanes, Now.

      13. Seattle drivers do not signal to change lanes, because they refuse ever to change lanes for any reason, no matter how irrational it may be to stay in the original lane.

        The only time Seattle drivers use their turn signals is when turning left or right, and they never do so until they’re about to make the turn, at which point they’re already blocking the straight lane and it’s too late to change lanes to go around them.

        Turn signals are for communication with other vehicles. If there’s no need to communicate (because no other drivers are near), then they’re superfluous. If you fail to communicate with them (by activating them too late), then you have done no good by activating them at all.

        Yet another example of Seattleites smugly extolling their perceived adherence to the rules, when in fact they’re making the city around them function objectively less well.

  7. Will these advancements make it to the google maps iphone app or just the desktop website?

  8. This is kind of cool – but most of the time I use the transit directions it’s because I want to be somewhere/leave somewhere at a specific time. I hope the ‘traditional’ option remains!

Comments are closed.