The thing that really matters in delivery is whether the address on the consignment has enough information for an operator to complete the next leg. By the time an item makes it to the region where the delivery address is situated, the local operators usually have enough understanding of the local system to get the item to its destination.
Even if the city in the article has a well defined system, it's probably not feasible for a global product like Google Maps to understand and encode every regional system. This is the problem that geocoding schemes (what3words, etc) are meant to solve, creating a single system that applies globally. But like many "rational" systems that attempt to replace entrenched practices, they struggle to gain traction.
https://gist.github.com/almereyda/85fa289bfc668777fe3619298b...
That a building|property will have only one address.
Sometimes (eg: rural Australia) property addresses are updated from an older numbered lot based system (that goes astray when properties are subdivided and infill houses appear) to a system that numbers houses by driveway distance from last major intersection.
For five or ten years a house can be recieving mail or be on the records with both the old and the new address.
It’s almost entirely impossible to order through Amazon et al using this type of system, it’s just not supported at all.
The same goes for my country or origin (in EU), they require my address in order to be able to send important mail. It’s just not possible because of the computer systems not accepting anything without a zipcode, address and house number.
James C. Scott would be proud.
That's not the original ethos of Google: organize the world's information and make it universally accessible and useful. I don't know about now but twenty years ago Google almost certainly thought it worthwhile to encode the rules of every regional system. Add that to Larry and Sergey's "healthy disregard for the impossible" I'm willing to bet that twenty years ago Google had almost certainly made it feasible to do just that: encode the rules of every regional system.
But I worked on Google's geocoding (mapping names to locations) and reverse geocoding (mapping locations to names) systems 15 years ago, and encoding every local ruleset was absolutely not how it worked. Or even encoding any of them.
And what's described in the post are exactly the kinds of problems you'd have back then as well.
Some of these were upstream data quality issues, some were due to deep infrastructure problems that could not have been addressed without a complete rewrite, and others just basic recall/precision tradeoffs around the value of returning a result that doesn't match the query exactly.
If my fixes had been published in the promised 24 hours then this blog post would not have been written but after two weeks this is the best idea I could come up with.
I think it is practical for Google Maps to understand the systems used in most major cities and then use this knowledge to reduce the number of errors.
I also think it is possible for the feedback system to work better. It does work sometimes, but it is slow and opaque and unreliable. It's even worse for bike directions.
The friend's cousin did what they could have done themselves, use the feedback tool.
Making sure buildings are near the street listed, in the right range of numbers, is a system that works in most regions and should be encoded and used for checking data.
Does it have to encode them all? Why not start out with one then at least Google Maps is a little better for some.
Besides that, how many systems could there be? There are only something like 10,000 cities on the planet. That sounds like the kind of task Google is built to handle.
I live in a small country of ~2mio pop, and we have approximately 6000 settlements with streets and street addresses and many different standards of numberings, depending on many reason, mostly historic.
Technically yes, a few thousand systems could be programmed into some google processing engine, but you'd have to manually classify every road to set the correct system, and even there, you'd never know what is a legit numbering scheme or what is an error.
For example, Cucumber street 1, 3, 5, 7, 9 on one side of the street and Cucumber street 2, 4, 6, 8, 10 on the other is a valid numbering scheme. If #3 splits his front yard and builds another house, you'd get "3a" on that side. If #6 buys #8, demolishes one or both buildings and builds a bigger one, that could become either a "6" or an "8", and the other would be missing. Then #7 is demolished and an apartment building is built there with 4 entrances to 4 separate building sections, so those would be numbered 7a, 7b, 7c and 7d, even though they're in the same building. #5, #7 and #9 are also a part of the same building (three entrances), but the building is older, the street was renamed and renumbered some time after, and each of the entrances got its own number. Then you come to #10, which is also a building with 4 entrances, but #10 goes towards Cucumber street, and the other three are facing the Lettuce street and are thus Lettuce street 6, 6a and 6b. Notice starting with the "6" here and "7a" above, well, that's because #6 lettuce street existed before the building was expanded, it kept the #6 number, but since there used to be a shed on Cucumber street #7, the new building starts with 7a.
Good luck writing a general model for that.
This is literally true when talking about the entire planet, but this is Vancouver.
There are many geocoding systems that do just fine with the considerable range of addressing schemes in the developed world.
Something very weird is going on here, like Google finding a cheaper method for geocoding that is probabilistic and yields a level of errors they are ok with.
You know who knows this? The tax authority.
> the local operators usually have enough understanding of the local system to get the item to its destination.
Large freight does not work this way. I worked on top of a ski resort for a while. We had a mapped address right where the ski left let out instead of the base of the hill. In the summer you could drive up to it on an ATV easily and a pickup truck if it hadn't recently rained. Somehow a semi truck driver for a freight company got this address and mapped a route to it. We were quite surprised to see the 53' box truck driving up the side of the ski hill.
> it's probably not feasible for a global product like Google Maps to understand and encode every regional system.
The information is encoded elsewhere and it's a bummer there is no incentive to make it as open and widely available as is possible. Although if delivery can rely only on partial information to "complete the next leg" then why can't address lookup do precisely the same thing?
I disagree, it's a requirement of being global. If you can't be global, don't. Meanwhile they have hundreds of thousands of engineers in their employ around the world, they have the scale to be correct.
I know this isn't what you said, but the idea that a small scrappy team of developers in one corner of the world can develop a system that works at all for the entire planet while dealing with the real world should be possible is just nonsensical. There's no reason Google can't be correct other than the fact it's unsexy to management that solving the problem involves hiring people who speak nearly every language on the planet and can connect and comply with enough governments around the world to get things right. Somehow that's not less surprising than sending cars everywhere to take photos of everyone's front doors without their consent.
Indeed... sometimes USPS even makes up their own "vanity city" names that don't exist. Or mailing addresses might use a different city name than where the house is physically located because of where the post office is physically located.
One year ago the Elizabeth line disappeared from the maps in London. There are many Reddit posts about this such as https://www.reddit.com/r/LondonUnderground/comments/1be01n3/... and https://www.reddit.com/r/LondonUnderground/comments/1b0xxb0/.... I asked a friend who worked at Google and they said that it was because some poor workers in India accidentally hid it while fixing something else.
It took google at least the best part of a month to fix it (at least, it was around a month from when I submitted the correction, no idea if others beat me to it)
Driving directions on the expressways here are also spotty. If I wait for Maps to tell me to exit the expressway, I'm already at or past the exit. I basically have to already kind of know where I'm going to make use of driving directions (you know, like we used to do with paper maps). Never mind that there are actual bugs in the directions: at one spot on the Ayer Rajah Expressway (AYE), if I need to exit onto the MCE, it tells me to exit onto the AYE, which is incorrect).
Still, I guess it's better than nothing.
If enough people using google maps have been down it in the right time frame it'll know the road is slow and account for it in time calculations, and generally not send you down it. The fact that no one using google maps has ever managed to go down that lane at 60mph never seems to enter into its consideration.
I've never had TomTom send me down a country lane unless it's literally the only way to get to where I'm going.
After two weeks of failing to fix the most significant error that I found I decided to blog about the issue in hopes of getting the attention of the Google Maps team, and also to share what I found.
There are 4 listed within 20 miles of my house that haven't existed for well over 100 years. They're not incorporated anymore, and they don't exist on any other maps, just Google.
It's weird.
It’s entirely visible if you overlay satellite with the map, but it’ll probably never be fixed.
Driving directions are a wonderful thing but they need to account for whether you are arriving in a ride-share vehicle (please drop me at the front entrance) or in a car you need to park (the front entrance may be worthless) - lots of work yet to be done.
My own first intuition is that it's not actually a data problem at all, and that "Google Maps has no concept of.." might simply reflect the ongoing, enshittening, transition from structured "concepts" to ML "vibes" for products like Google's.
It's not that the underlying maps data store has the addresses wrong, but there's a layer between the input field and the result generator that's statistically deciding you mean something besides what you explicitly enterred and is giving you a route to a silently "corrected" address.
We've seen that happening more and more in Search for years, silently ignoring keywords and directives without a "did you mean" callout, and it would seem natural for some product owner to be pushing an equivalent initiative in Maps. Aggregate metrics move the right way, so the company is happy, but of course the actual product experience sees a fractal failure pattern that nobody can quite address but makes results less and less reliable.
I'd love to be wrong, though, as I think this would be a terrible advent for something that can be life-critical like a popular mapping tool.
In particular it was developed in the western world from a startup, so the first approach probably was minimal and adapted to local streets, without even covering the edge cases.
Going from that to mapping every single address in the world is a huge leap and the underlying data system must be an incredible mess, also creating regressions on what used to be reliable.
I don't know if they got to it, but a few years ago you couldn't have vertical stacking (e.g. floors), shops with the same address needed workarounds to have a different entry.
Then some places have multiple addresses. The article talks about street numbers making sense, but in most places in the world they don't, disappear at random points, some countries don't have street names. We could talk about it years.
The "city name" on an address isn't really a "city". SFO's address is "San Francisco, CA", but is not within SF city limits.
Queens NY addresses have "cities" that are just neighborhoods.
Applying any kind of logic to addresses will just be a minefield.
Apple Maps drives me nuts—it will only return search results if you include the hyphen in the four digit street address ie: "36-08 33rd St" vs "3608 33rd St". Google will hit on either query.
The hyphen is a part of the "official" address. However, USPS has declared it unnecessary, there's no advantage to using it unless you're navigating by analog map, and it's a PITA to type on a mobile keyboard.
So if anyone on Apple Maps team is here: please fix this. I filled a apple.com/feedback ticket on this years ago.
For anyone interested in the peculiar history of Queens addresses—they convey a cross street and the number of the house ascending going northward.
For example: 36-08 33rd St means that the house is on 33rd St, between 35th and 36th ave, and is house #08 on the block.
https://stevemorse.org/census/changes/QueensFormat.htm https://www.nydailynews.com/2011/08/21/balderdash-queens-res...
That said, Apple Maps is far superior to Google Maps for transit directions, at least in NYC. Google's integration with the MTA is seriously lacking—their directions often do not reflect scheduled changes in routes, let alone real-time issues. That said, Google Maps is superior with POI search and address decoding.
The fact it gets the street wrong indicates a tokenization issue somewhere.
Any evidence it happens to non-numeric streets like Main Street or Martin Luther Blvd, or is it only 10th st types of things?
Holy shit what an absolutely needless nightmare of a system. How do you expect people to find anything like that?
- Municipalities are divided hierarchically into named neighborhoods, districts, wards, or "sections" (字/"aza", I don't know the correct translation). A single municipality might combine different kinds of divisions for historical reasons.
- Those are then divided into numbered blocks, plots, and buildings. The numbers are assigned in chronological order, and are rarely spatially sequential.
All in all, the system is extremely confusing, and the chaotic urban "planning" of most Japanese cities doesn't help. So how do people move around? They use maps; in the big cities like Tokyo there are street maps in all the major areas, otherwise you use your smartphone.
In some areas the streets have no names, nor do the buildings have numbers.
If there’s three streets and twelve houses, why do you need any of those?
Sometimes I dream of going back, but the culture has changed too much (and not for the better, I hear).
So we now regularly get people driving down here or trying to park, when it's actually just a one-lane driveway.
I had a problem where Google Maps showed the wrong location for my address. I reported the issue. I was told the correction was accepted. But it wasn't — Google Maps would still show the wrong location. After doing this dance several times over a course of weeks going into months, it would sometimes show the correct location, and sometimes the wrong one. I once had two people arrive in the same car, and each of their phones showed a different location in Google Maps.
The problem is that at this point Google Maps is so ubiquitous that people accept it as "truth". And that is a problem when it shows an incorrect location: people can't get to your place, package deliveries are delayed, etc.
Unfortunately, things that are so ubiquitous in our lives require regulation — at the very least, Google should be required to process (and verify) requests for corrections in a timely manner.
Sometimes, the local addressing authority was wrong, and I would have to prove it to them. "But it was checked!" It was checked wrong. Street numbers would be off by thousands. It would take some pointing out of obvious problems in the progression of numbers, plus a plat, plus an email from the building manager to prove to them that I knew what I was talking about.
I was contacted by a woman who kept having drivers for all kinds of services attempt to use her driveway as a street, even if she had a sign up, which she did. Her local municipality was no help. Nobody was. She was irked and frantic, as these trucks would destroy her driveway, lawn, even garden.
Much digging ensued. It turned out that the proximal source of the error was a statewide system, one of many which Google simply hoovers up and digests like a baleen whale siphoning up plankton, then digests and tries to "make work." I got the proximal source to make a correction, which they might publish in another three months. The original source of the error was a long-missing minor street from many decades ago, which I had to find in caches of searches and the like.
It was idle curiosity, but it took me about twenty hours of digging on evenings and weekends.
Address points are easy. Parcels, aka the polygons upon which zero to many address points might rest, are harder. Road networks are terrifically hard. I managed to catalog ten different cases of road discontinuity in the process of trying to find such things in an automatic fashion. And I might not be bright enough to have enumerated them all!
And then we have cases of people who deliberately insist that their address number is "00." As in the first two-thirds of a certain not-so-secret agent's code number. Or imagine the fools who decide to put up a sign at the end of their long driveway and simply declare that it was a road.
Each county has its own addressing standards, and included in each are addresses from the Olden Days, real wild west stuff, which the authorities are just itching to scrape out of their systems once and for all.
Road addressing is Fractally Bad.
https://duckduckgo.com/?q=136+W+6th+Ave%2C+Vancouver%2C+BC%2...
It was Google that did not bother to update their data for years.
I've had this problem for the 10 years I've lived at my address. Google has even changed where they usually send people, which is just a different kind of wrong. Non-existent locations on the other side of town.
It's very frustrating. I usually give people the address of a nearby business. I dare not order food delivery.
Once I ordered a consumer product "next day delivery" which I had no idea would summon an Uber Eats driver. I texted him the correct location once he showed up to the incorrect one. When he got to my house Uber wouldn't even let him complete the delivery because his GPS was so far from what it thought the delivery location was. He had to call them to resolve it.
I've tried to correct it myself about 10 times. I have had all my friends try to correct it from their accounts.
I still don't think that Google Maps should be regulated though. I think it's a ridiculous thing to suggest. I just don't think regulation is the answer to every inconvenience.
Openstreetmap already has the correct house outlines and the new street.
I already warned them that deliveries might be tricky for a few weeks/months.
Addressing is hard and errors are common. You can't blindly rely on maps to be correct or up to date.
Confusingly, for example, Google thinks it sometimes takes like 20 minutes for the metra electric to go from Van Buren to millennium stations (a distance of perhaps half a mile). What I believe is happening is it's using the departure time of the train from millennium station (it will layover for 15 minutes or so, presumably) as the arrival time.
Waymo is convinced that it is impossible for a pedestrian to walk from where I live to the east side of my lot (I live in multifamily, 300 units).
Waymo seems to have pedestrian pathing in mind, and will block the app from a walk that is too far away.
The landlady provides custom made pickup areas in the east lot, but Waymo keeps insisting on pickup and dropoff on the west side.
Adding insult to injury, the “set on map” feature of dragging the pin has been broken as well.
So I have to put up with it. I mean, they’re still in “testing phase” but I’ve given them such clear feedback, they are surely aware by now.
Because of that, I ended up building a tool called AddressHub that focuses mostly on cleaning up the input before sending it to geocoders. It tries to remove instructions, phone numbers, weird formatting, and fixes common spelling mistakes or street name variations. Then it sends the cleaned address through multiple geocoders, checks how close the results are to where they should be, and picks the best one.
It also caches results when possible (depending on the API terms) and tries open data first to save costs. Only if open data does not have a good result, it falls back to the paid geocoders.
This has helped a lot with reducing wrong deliveries caused by addresses that look fine at first glance but are actually way off. If anyone’s interested, the project is at www.address-hub.com. Always happy to get feedback if people have ideas for making it better.
This was particularly weird because normally the geocoder was fine with bogus unit numbers (eg on obvious single-family homes) but fell over in just this one case...
https://streeteasy.com/blog/queens-addresses-hyphenated-conf...
The only consistent thing about addresses is that they're really not that consistent...
It's not only Google that has this issue; but the nature of addressing in general. It's not uncommon for the geocoders we use to be hundreds of metres off of the actual location of the address.
The addresses there are notoriously local-centric, with knowledge of the area required, to find a physical location from the address.
All that being said, I wonder how long it will take people to get the memo: Don't use Goggle! For anything! Unless you like being surveilled (there's a fetish for everything, I guess)
I've used osmand on a (community build) android phone for about 8 years. It works great offline, all maps and the nav algorithm are in the phone, and it's gotten me to where I was going.
I'm hoping to migrate full time to a linux phone this year. The mega-corps have contributed to the long wait for this tech, but it does seem on the verge of finally breaking out of the lab.
Before that, it would give reasonable mass transit directions in San Francisco. After that, it would always use a bus to go even 1 block to BART, giving drastically inflated transit times.
This smells like some weird caching, although I do most of my tests in an incognito window to minimize this.
And I have no idea what to do about this.
P.S. I never had good experience with reporting errors. Sometimes they are rejected despite absolutely being correct, or they end up having no effect at all.
The response from GM was that they did an investigation and the address is correct. I replied to the dude in whatever country but not mine that I bike there daily and they even have a clean picture on street view.
Then suddenly after a year, or if the blue, I got a Thanks To You It Is Fixed automated email.
This is a weird company.
I've always wondered who reviews the requests, if you are a local, or even spend 2 minutes checking, you'll notice the mistake...
Whatever assumption you think will always hold true for addresses, will not in practice. Even in places that appear strictly patterned.
The suggestion to guess addresses that "should exist" is just clearly wrong. If someone buys two plots of land and builds a larger building, an address normally vanishes.
They don’t need to understand how addresses work, just trust that most users have put in their home address correctly.
If we’re going to have all our data siphoned up and aggregated it would be nice if there were some useful side benefits.
Surely that's a typo and supposed to be 12.3 furlongs? Even that might be slightly incorrect.
I don't envy developers that have to work with geography or timezones. And Google Map devs have to work with both. Ouch.
What I really I want are options between where I am to where I might be in 30 minutes.
Google Maps is really bad when streets are even slightly unusual. That's even with North American address conventions, which are extremely regular by comparison to many nations.
For example, if you have both a North and a South version of a road, and the address numbering for North and South versions overlaps -- say, North and South begin at a county midline road and North increases going north and South increases going south -- it will irregularly put a pin on the wrong road or wrong segment of road.
Similarly, it will be confused if you have River Rd and West River Rd. And it can be confused when Oak Ln becomes Oak Ct or insist that Oak Ct is really Oak Ln even when the road name actually changes.
It also gets very confused by roads that have two names. For example if there's a county line road, the east county might call it "Franklin Rd" while the west county calls it "E County Line Rd." And in that case the east side of the road have one set of addresses, and the west have another set. Worse, the address numbers often don't align. Except that's not what Google screws up that often. Instead, Google will sometimes insist that one or the other road doesn't exist at all. It will say that 123 E County Line Rd is actually 123 Franklin Rd, and then it will put a pin where 123 would be on Franklin Rd if that road didn't begin its address numbering at 3000.
Sometimes it insists the city is incorrect, too. If your address is "123 Miller Rd, New London" and New London is a tiny unincorporated town near Portland, Google might translate the name to "123 Miller Rd, Portland." Sometimes even when there's a street the next county over with an address "123 Miller Rd, Portland". In this case if you enter the Portland address, it will point you to the address actually in Portland, and if you put in the New London address, it will show you the New London address... but it will still correct your New London address to Portland.
If you have a road with breaks in it, such as for a river without a bridge, it will occasionally just... put a pin at the end of one segment of the road and not find the address on the correct segment on the far side of the break.
About the only things that's really consistent is: If you zoom in and the pin is in the middle of the road, then Google Maps probably can't find the address. On the other hand, if the pin is off the road, then it's probably exactly on the structure based on the local or municipal authority and their GIS data. In that case, Google found the address on the GIS data they got from that municipality.
And you might say, "Oh, but those are really easily confusing things! It's entirely understandable." And, maybe that's true. But USPS's ZIP code finder still knows the addresses well enough to both find and correct them for you, and ArcGIS interfaces also seem to be able to find things much more easily. Google Maps was groundbreaking 20 years ago. But it really hasn't kept pace. The only thing that seems to confuse the other sites is new construction. At the very least, I wish Google Maps would be more clear when it's guessing rather than when it's found an exact match.
All that is to say, yes, we did use Google Maps to help find addresses. But when things looked even a little weird, we assumed that Google Maps was wrong. And it usually was wrong in those cases. And what it got wrong was sometimes really, really wrong.
Soon we can tell the younger generations that when we were young, Google was almost magic, it could find the most obscure stuff.. nowadays, not so much.
Should be "the writer of this article has no idea how street addresses actually work in the real world".
