A recent post discussed Jeff Howard's hotels near transit maps suggested looking at Google Hotel Finder, a utility tucked away within Google Maps that purports to help you find a hotel based on travel time from some location. A user plops a pin on the map, and the tool draws isochrones based on drive, transit, and walk times, which supposedly show you the area of the city where hotels are within that travel time of your destination.
So far, so good – I put a pin in downtown Portland, and Hotel Finder shows me a big blob in the center of the city that I can get to by transit within 15 minutes. It looks like there will be lots of hotels I can choose where I can quickly take transit into downtown. However, when we look a bit more closely, we can see a big problem with Google's approach.
In this image, I have moved the pin to a corner near Reed College in Southeast Portland. This is a residential area bordering a low-density industrial district and the campus and fields of a small, exclusive liberal arts college. It is served by just one bus route, the 10-Harold, every 30 minutes. Yet according to Google, from this location, I'm just a short 15-minute trip from outer East Portland, or the inner Eastside Industrial District.
The problem here is that Google is providing an isochrone of transit access that does not consider frequency, i.e waiting time. They assume that the bus shows up right when you need it..
Once I'm on the Harold bus, it's true that I might be able to take it from 28th far out into the east side in just 15 minutes. But depending upon when I arrived at the stop, I could wind up spending up to 45 minutes making the trip. If we assume an average wait of 15 minutes, or half the headway, the area shown by Google as within 15 minutes of the pin is actually more like 30 minutes!
Imagine you are a person who is coming to a city for business, and you picked a hotel expecting to be able to travel to your meeting by transit in just 15 minutes. Yet when you walk out to the stop, or check a trip planing app, you find that you will wait longer than that just to catch the next bus! You might be late to your meeting, and the tool you used to pick the hotel would have failed to direct you to accommodations that met your desire to be a short transit trip from work.
A more useful version of Hotel Finder would add waiting time. This would alter the isochrone in response to frequency, and more accurately show the area (and hotels) within a short transit trip of the desired location.
We are surprised to see this kind of misleading info from the crack team at Google Transit. In presenting transit travel times that don't consider waiting, they are talking about transit as though it worked just like cars, doing a disservice to everyone who wants to consider transit when the choose a location.
It also looks like they’re using circular buffers around reachable transit stops instead of network buffers (actually following the road network). For very small buffer areas, it doesn’t make a huge difference, but for larger areas, the circle can drastically overestimate the area that’s actually reachable if you’re constrained to walking along streets.
Melinda. The buffering problem is tough because it requires a model of the pedestrian network, which isn’t routinely available in most cities. Also, if you use a walk buffer you need a longer threshold distance. See Chapter 5 of my book.
Jarrett: by total coincidence, I’m working on a startup which will address the hotel/transit problem you describe. We’re making a travel-accommodation search engine that is built around the idea that being close to the things you want to see and do is essential; it also takes transit stations and preferred transport mode into consideration. We’ll be launching our beta site later this month, and we’ll be factoring transit frequency into our calculations soon. More info at http://www.nexmoov.com.
Google still hasn’t fixed the problem with their driving directions completely ignoring the problem of finding parking. They show you drive time from front door to front door, but don’t give any guidance on whether you can park anywhere near there, and how long it might take, so they often mislead people into thinking that it’s quicker to drive somewhere than bike or take transit, when parking would reverse things.
The fact that they don’t address this central concern with their central navigation product suggests that they’re unlikely to fix problems with the transit navigation that are much more remote from their experience.
Google has a model of the road network, and often has peedesrian-only connection too. They certainly have enought data to do a lot better than simple circular buffers
Google’s transit directions are oriented around providing discrete trips — identifying the shortest possible timings available to fit a specific time window. This makes sense in a narrow way, but it does make it difficult to determine what is generally the best transit option, or what a practical ballpark travel time is. Naturally it’s terrible for the purposes of the Hotel Finder app, because that’s not what it’s designed to do. Definitely something they should work on, though.
The fact that this particular feature doesn’t work properly doesn’t mean that Google Maps isn’t a usable tool for finding hotels near transit.
Google Maps will show you hotels, and it will show you transit, and a bit of clicking around will let you ascertain details about said transit. With a bit of eye-balling and common sense, this will work fine for many cities. A visual overview of the problem-space is very useful, and that’s what Maps gives you.
Granted it probably works better for train networks than for sparse infrequent bus networks, but the latter tend not to be very good for tourists anyway, and if that’s all a city has, you’re probably better off just choosing some hotel that’s relatively central and crossing your fingers…
This problem is prevalent across all trip planners, as it is impossible to know an exact departure time for these maps. As Miles points out, frequent transit does work better, but a compromise for these maps could be showing a minimum and a maximum trip time for lower frequency transit. A similar issue also appears with cars in traffic, as the accuracy of the predictions is only as good as the historical data, or the accuracy of the current travel time.
Ah found this post with this text I remember …
“We are surprised to see this kind of misleading info from the crack team at Google Transit. In presenting transit travel times that don’t consider waiting, they are talking about transit as though it worked just like cars, doing a disservice to everyone who wants to consider transit when the choose a location.”
It seems the guys at Google Maps heard you. I am not sure when but it seems they now include “waiting time” equal to “frequency of service” at the start of a proposed trip. For example this Google Maps Transit Directions query https://goo.gl/3WsUuz shows a wait time of 20 minutes at the start of the T203 bus journey (frequency 20 minutes at the search time). Screenshot here https://goo.gl/photos/BPqMceJjFrDgJJMCA .
It is really a good article as the idea depicted in it, is very helpful in finding the best hotels in Derby town Centre. You have described the fact that all the information that are shown on the google are not true everytime, you should also use your own mind.
Very helpful and Great information,
we appreciate advise especially coming from a professional.
Thanks again and keep up the great work!
For hotels and motels Read more at http://vibrantwave.com/cheap-hotels-in-new-york/
Really so good – I put a pin in Derby, and Hotel Finder shows me a big blob in the center of the city that I can get to by transit within 15 minutes. It looks like there will be lots of hotels I can choose where I can quickly take transit into Derby City Centre.
Google has a model of the road network, and often has peedesrian-only connection too. They certainly have enought data to do a lot better than simple circular buffers.