Once again, HN is the last-resort customer support line for Google. Oh well, upvoted. Embarrassing the company is the only way to get any slightly complex issue resolved.
From the outside, you just can't imagine how few Google engineers are responsible for how many products and data integrations.
One example: back then it was a dozen people (literally) responsible for every single business's hours, category, and more for all of planet Earth. They automate the crap out of it and the automation mostly works (scrape structured data off websites, Google My Business, various aggregators, etc.) but it means there's no conceivable way you could provide customer support that will keep up with the volume of support requests because the support requests often ultimately have to get routed to an engineer... and there are only a dozen engineers on the team that owns stuffing all this data into Maps. And those engineers are supposed to be doing mostly feature work. Over time accuracy got better, but 95% accuracy, to say nothing of 99.9% accuracy, is unachievable to the point of being a non-goal.
One of this team's bigger time sucks was just other Google engineers complaining about inaccurate data. Someone would go to their favorite restaurant, notice the hours were off, or the business category was wrong, and file a ticket. Then the team would sometimes manually edit the listing to fix it. Sometimes they'd say "no, you have to deal with the same shitty data as all our customers". But that was just the requests coming in from a few thousand users who had access to the internal ticket system and new which component to file stuff under. (Incidentally: this is the origin of the Maps feature where you can now just directly edit a listing and if other users, agree, it gets published.)
So, yeah, there isn't really a scalable way to manage support requests. But HN works for some subset of cases, and I too upvote the cases that get complained about here.
EDIT: accuracy in the raw data was pretty bad but on a view-weighted basis it was pretty respectable IIRC (>90%). Turns out most people on Maps are looking at the same small subset of establishments, so you have to get those right even if that means making choices that reduce accuracy on other listings.
I mean, I've seen understaffed, shitty-service companies since I was a kid. How is this unique? Fuck them all.
Democratizing a thing mainly involves scaling it, and scaling things rapidly is a difficult task. Having only a few qualified people ready for a task of this magnitude is not surprising to me.
So what you're saying is, there is this hugely valuable application that has "revolutionized" things for a large fraction of the population, but it's perfectly normal that Google can't be bothered to put more than a piddly number of people on it?
If it's really that valuable, Google should not handle it in this half-assed fashion. They should charge money for it, just like any other business that has something it knows is hugely valuable to a large number of people. But instead, they prefer to sell everyone's eyeballs to advertisers. How does this make sense?
Google used to run a program called Contributor which let you pay them the minimum amount you'd have to to opt out of ads (you basically just paid them and they used your money to automatically outbid everyone who wanted to advertise to you). I wish they brought this back because I really liked it. They changed it completely so it only worked with "participating sites" [0], which was basically nobody, then killed it altogether (but left the page up).
There's an easy solution to this: what economists call price discrimination. Basically, people who can afford to pay more, pay more, and people who can't afford to pay more, don't.
It looks like the Contributor program you describe could have been set up to do this.
Yes, I get that Google and Facebook and Twitter and friends have gotten people accustomed to having all these wonderful services that everyone loves to use somehow magically appear on the Internet for free. That doesn't mean it's actually free. It just means that the costs are hidden.
Which is all fine in and of itself: if everyone else wants to hand over their private data to be sold, sure, go ahead, do it. I don't even have a Twitter account, and while I have a Facebook account I almost never use it and have entered zero data beyond my name into its profile. I can't quite say the same about my Google account, since I do buy books and apps from Google Play, but it's as close as I can get it to zero data in its profile. So I at least have some way of limiting how much I am participating in the great selling of personal data.
However, I don't have a way of getting better service from Google for things that could be better services if those of us who want the better service were allowed to pay for it. That's one of the standard ways of doing price discrimination, and should be very familiar to anyone who has dealt with cloud services, or indeed anyone who has used an app with a freemium business model: service tiers. The lowest tier is the free one; the other tiers are progressively more expensive and provide more added functionality. But Google can't be bothered to do it for its hugely useful apps like Maps and search, and then they tell us how difficult it is to staff these hugely useful apps with enough engineering resources. It doesn't add up.
I'm going to say this as a user. Google Maps is fucking fantastic. The team responsible for it should be proud:
* Free
* Great data - almost always accurate
* Editable - I can contribute
Genuinely an amazing product. And all these people on the Internet saying it's shite. No fucking way. It's fantastic. If you ever listen to these people for even a second, go and listen for half that time to folks on the street. They love your shit.
I am with you on not being willing to pay anything for the service.
They are also complaining that someone wants to run a public utility and be the centralized point for information but also doesn't want to pay enough to make that public utility accurate.
Not saying that its even feasible to do so, but that's why we pool funds for utilities - we all use them and we all have an incentive to make them not shit.
What if it were not offered for free? What if you had to pay a subscription of, say, $10 a year? Would you pay for it then?
If your answer is yes, then it's not that you're not willing to pay anything for it; you are simply taking advantage of the fact that right now you don't have to. But the fact that you don't have to is a side effect of the ad-supported business model, which IMO is doing a lot of damage that could be avoided if only Google would take the simple step that businesses have taken since time immemorial, of charging its users directly for the services they use.
If your answer is no, then evidently this application is not as great and wonderful and useful as people are claiming it is, since it isn't even worth a $10/year subscription, i.e., about 1/10 the price of Netflix.
Google Maps is universal, dude. The fact that you can't even comprehend $10 being a lot of money is amazing. Truly truly amazing. But I must say I fully expected it.
The fact that you can't even comprehend that to anyone who is posting here on HN, $10 is not very likely to be a lot of money, is what I find truly amazing. Not to mention your blithe acceptance of something like Google Maps being "universal, dude", as though it were a law of physics that such complicated and useful applications will just appear for free, or that there couldn't possibly be any downsides to such a state of affairs.
"Almost always" when you have billions of data points means millions, or even tens of millions, of wrong data points. That doesn't seem great to me. It certainly won't seem great to all the people and businesses whose data is wrong and who have no way of getting it corrected. Which is exactly the point of the article this discussion is about.
> all these people on the Internet saying it's shite
We are saying no such thing. I agree that it is a hugely useful application. I just think it would be even more useful, not to mention less damaging to all the people and businesses that are hit by the "almost" above, if it had at least some of its users as paying customers instead of being solely driven by Google's ad-supported business model.
I understand the idea of "automate it, it must scale", but especially then I'd expect the systems to add/correct/delete info to be super streamlined, the documentation to be very thorough and covering all business cases etc. Yet it feels like the opposite is true: the GUIs are generally sub-par and feel like "okay, we have a great API. Can you throw some GUI on it?", the documentation is short and often either outdated or did not cover certain cases in the first place, and "what to do if this doesn't solve your problem" is rare.
Is that "we want to be a platform, not a service provider" and meant to discourage end-users from using the system directly and pay a professional agency instead that handles the support issues? Because that's really the only explanation I've heard so far that makes some sense. It feels like Google is putting more manpower in developer relations than in customer relations and I've never understood why.
I've seen dozens of articles on this site about hiring practices from large technical companies, and Google in particular.
I don't remember seeing anything suggesting they even verify that the people they're hiring are capable of writing basic documentation, let alone that they appreciate the importance of it being complete, correct, and up to date.
Every indication is that Google vastly undervalues documentation, both for internal and external purposes.
This just seems to be companies as a whole. Documentation is a low-status activity (no promotions or raises available for doing it) that still requires decent cognitive function, so it is not something you want to do when you are energized and ready to go nor something you can do when tired and not in the mood to work.
There are no automated tests to tell you that the documentation no longer works, so the engineers would need to periodically check. We have automated tests in code because engineers won't comprehensively check everything.
In my org we frantically document when an engineer gives notice. Otherwise, we can't even keep the URLs updated or maintain a running list of configuration variables.
Digging yourself out of the hole where your documentation is already wrong or incomplete is much harder: when you go to describe a change you might find that the thing you're changing isn't mentioned, or is described wrongly, or there is no terminology for the thing you need to talk about.
So you need everyone to take care that standards don't slip, which is why it's important to make the importance of documentation part of the company culture, and make sure every programmer is capable of doing their bit.
He didn’t get a promotion, but he did have a lot of positive visibility. When there was a round of layoffs, other teams lost engineers but we got to keep him.
Large companies hire technical writing specialists to do documentation, not software developers.
Did you think to check whether the new hire who can write a regex parser in his head can also design a library that other people want to use without screaming?
I updated some opening hours and needed to delete a dead link on the search engine. All this stuff is automated and works great.
Unfortunately this means there are still something like half a million edge cases that get issues like unwarranted account lockouts.
The Maps feature for editing listings was a ton of work spanning a bunch of teams over several months, but that kind of investment is made only after every other option has been exhausted because it's so slow and expensive and scales poorly relative to just ingesting more data and throwing ML at it.
E.g. the Knowledge Boxes on SERPs. When I was invited to add/edit to them for a notable client a few months back, it was barely documented, cumbersome, error and response messages were vague and it felt like a proof of concept.
I guess I'm just wondering: is that by design, has someone tested that and figured out "if people feel like they're using ebay's dysfunctional interface, they'll spend more money on ads", or is it just that somebody gets a task "hey, add a form to edit that, please" and they're like "well, they didn't ask me to make it pretty and I really need to leave soon"?
I saw some A/B testing for ad engagement that drove UI decision making, but it was always short-term and related to search -- like, "we updated the model today, and users click on the ad 1% more often in the treatment group".
FWIW I was pretty much at the bottom of the totem pole so if there was some nefarious plot to provide crappy UIs I wouldn't've been in the loop anyway.
it sounds like there needed to be (needs to be?) tooling to allow customer support to make these changes. Fixing these kinds of things really shouldn't be something that engineers should have to do manually.
The problem is that (a) business can mistakenly enter wrong data (b) businesses can lie (c) businesses can fail to enter any data (d) businesses can fail to update data.
Google can't let any random idiot with a smartphone edit Google Maps, and they also can't let any random idiot waste a potentially unlimited amount of customer support time on what might well be an erroneous change. So they came up with a thing where you can edit Maps but the edits only apply if some conditions that can be automatically checked are satisfied (recency, user edit quality track record, agreement among users, stuff like that). But building and validating that was a big project spanning quite a few teams and several months. The opportunity cost often isn't worth it for that (I don't know whether it was in this instance, either... but it is really cool IMO.)
This wouldn't matter if Google weren't a de facto monopoly.
If Google were forced to spin Maps out such that it had to stand alone or die rather than being subsidized by the Eye of Sauron, these problems would begin disappearing.
Yes, because businesses struggling for cash invest in support for people who aren't paying them...
Google gets away with the quality level of their services because they are the information discovery monopolist.
I think those two services are incredibly strong products and gmail is additionally a positive equalizing force, allowing much closer to everyone to participate in modern communications.
It’s maybe technically true that these problems would disappear if maps had to feed itself, but I think that would be a net negative for the world.
Hard to say for Maps. There were maps freely available before Google Maps, but they didn't look as nice and they didn't have Streetview and the routing services, and you couldn't embed them on your Website for free (well, paying with you visitor's data, really).
For Gmail? I feel like emails were more reliable before Google decided that any email is likely Spam unless it's sent by one of a dozen commercial providers. Gmail didn't jump ahead in features like GMaps did vs the existing systems, it had a nice web interface and came with a lot of free space in exchange for reading your emails to create a better profile for targeted ads, but that was pretty much it.
I used to have no spam filtering at all on my mail (other than human visual grep). That was ok through 2003 and barely tenable up until about 2007. I then tried to run my own with SpamAssassin and the like. Around 2012 I folded and switched to running inbound mail through gmail. Life immediately got massively better. I almost never see spam and my own domain mail (outbound self-hosted, no commercial provider) gets through to gmail users just fine.
Yes, I get some false positives (ham categorized as spam). I probably have undetected false positives as well. It’s still 100% worth it as email was getting difficult to use previously.
Google false positives on the other hand are a pain in the butt. Even sending from one GSuite account to another on the same domain does not ensure delivery (but makes it much more likely than using a dedicated server that has low traffic to Gmail). I've adapted and now use my own Gmail account if I need to send email to Gmail users that is somewhat important.
Yes, because "free" collapses that entire market and makes it hard to ever get anything else, all while granting a massive amount of power to a company that is happy to break the ecosystem. GMail 1. makes it hard for any new email companies to get off the ground by making it hard to get income, while 2. making it hard for anyone (not an established major player) to successfully send email without it getting blackholed. That is, I blame GMail for the fact that I can't spin up my own mail server and send email and have people actually get it.
Perhaps put better: GMail and Google Maps did improve on what was then available, but in a way that also makes them dead ends for the entire industry, unable to improve beyond a certain point because of how they work while at the same time preventing anyone else from being able to make something better.
Got a typo in a business listing, wrong opening hours, &c.? Sure, you can fix that.
Got a road segment slightly wrong? OK, you can straighten it out.
Any other case where the underlying map data or presentation is wrong? Nope, you can’t report it, and any attempt to report the issue by the channels that are available will fail to have any effect.
And that’s the problem, Maps just doesn’t seem to be interested in providing, y’know, maps. It’s focused on being a business directory instead.
Take https://www.google.com/maps/place/Navarre+VIC+3384,+Australi... as an example: that pin and label on the map is almost 5km south of Navarre, because they seem to have taken the entire suburb area (which in rural areas is regularly many kilometres across) and picked a point on its boundary—I dunno, maybe the furthest point south, maybe just the start position of the polyline—rather than taking a sane point that represents where the town actually is (e.g. its post office) or even a midpoint in the suburb area.
When you look through rural areas you find a substantial fraction of the labels are wrong by several kilometres. Head south to the next town, Landsborough, and it’s marked 4km down the road from where it actually is. I’ve seen some places 10km off the mark.
I’ve attempted to report the Navarre case six or seven times over the last few years while living there, by three different ways (none of which is properly applicable), and it’s still stubbornly in the wrong place.
(Yes, I’m mentioning this here on the off chance that someone relevant will see it and pay attention.)
(That particular issue is partly resolved by OSM data changes that have happened since: it’s still using the administrative boundary, which is the wrong thing for it to do, but seven months ago someone added an admin_center to that boundary so that Qwant Maps drops the pin on the right point rather than the centre of the boundary—though it still zooms to include the whole boundary, which is much wider than it should zoom.)
On top of just mistakes there will be active fraud
What differences between Wikipedia and OSM do you think prevent OSM from becoming the best map?
And why is that? Google is not a poor company, can't they add a separate data support team for Maps which is a pretty important application? Is Google that cheap?
Why should feature engineers work on data support? I understand they implement automatic validation for data, but in addition to that there should be a separate data team which checks data manually.
The idea of needing to roll out, manage, moderate huge teams of people to manually review data and make judgment calls would mean needing to take responsibility for the listings and needing to care about micro-accuracy. Google doesn't want to be the "documentation team of the world," they want to be the "automated documentation generator for the world." Because that's an interesting engineering problem. As soon as it becomes primarily a people problem, Google individuals would just rather work on other things all things considered.
Even if there is a business case for making this information more accurate, it's just a fundamentally different mindset. You can't hire your way into getting that work delegated very easily either when the mindset of "solve it once with code so that you don't need to solve it 5 billion times by hand" permeates the culture.
Well, they are startup, low on cash so we should understand. As they start making money, no doubt they'll step up. /s
Google should shunned by us and investigated for antitrust.
I mean, from one perspective, sure, it'd be nice if there were a team of thousands to handle this problem.
But from another perspective, you just said there were a dozen people staffed to handle a single feature of a single product. Out of how many features, out of how many products?
My first job was at a start up back in the comparison shopping era where our goal was to get every product for sale on the internet into one place. For several years, I was responsible for all of indexing and search; another engineer was responsible for everything related to the front-end, another for everything related to crawling and scraping, and a forth worked on a variety of projects. Four engineers to cover every store's inventory in the world.
So, you know, coming from that world, the idea that there are a dozen people responsible for just business hours is crazy.
(For completeness, our engineering team fluctuated as high as a dozen when the company was flush with investor cash, down to four for a long stretch when the company went bankrupt for awhile, and then up to a half dozen or so when the company was profitable for a span.)
If we looked at a breakdown of whose info was accurate and whose was incorrect, there'd probably be some harmful biases in there, is that right? Isn't there some sense of responsibility here? Like, it's understandable how difficult it is to do this in a way where there aren't a small fraction of people being harmed, but that's not really a defense that they aren't causing harm. It seems like a defense along the lines of "It's impossible to be the good guys at scale for free, so of course we can't be the good guys."
You have to pick some balance of precision and recall. The yellow pages and paper maps did the same thing. They were handmade so they aimed higher on precision and lower on recency/recall. (Imagine opening a new business the day after the Yellow Pages has gotten dropped off at everyone's front door. You're screwed for the next year or whatever.) Maps hasn't picked the wrong balance; it's just a different balance that as a user you have to understand and account for.
I suggested a edit to revert the “closed today” hours and sent a picture of the hours. The edit was somehow rejected immediately.
As a “local guide” who volunteers to improve listings of local businesses, this left a sour taste in my mouth.
I also don’t believe that out of the 50 Googlers who were there that day, not one tried correcting the “error”. This leads me to believe it was malicious.
If Google was a scrappy startup with a few dozen employees, people might be more sympathetic. But Google has hundreds of billions of dollars at its disposal to fix the problem. It could purchase a small country and employ every person in that country to fix its problems. But Google won't admit that people can do what algorithms can't, and that's what causes Google to get so much hatred from the public.
And this is for positions that would increase the run rate or help make a business more profitable! But when it comes to businesses that are not profit centers - like maintaining business listings - I would expect those to run pretty lean. Google isn't a charity, after all.
As a happily employed engineer who has FAANG recruiter friends and gets scouted fairly regularly, I think you also forgot "and is willing to jump through the hoops you put in front of them".
Also, this doesn't sound like it needs to be an engineering problem. A customer support team could alleviate the workload. Even if its an internal-only, non-public support team, just people whose job it is to do the manual data scrubbing.
This sounds like a possible revenue opportunity as well. People would probably pay for the ability to ensure accurate data for their Places, which could also serve as a way for G to ensure the Place data is valid and meets their editorial guidelines. Don't know if this already exists or not but it sounds like it should.
Paying people to manually scrub data is possible, but extremely expensive at scale. I assume that the folks in charge of the business are looking at the numbers very carefully to see what justifies that kind of very expensive personnel investment and what doesn't. Also, I suspect that a lot of people will be angry that they have to pay to correct Google's mistakes, especially if their competitors don't. That's not the kind of business model I'd want to be involved in.
And yet thousands of companies, large and small, have done it for at least a century, even back to the times of paper records.
Google doesn't get a pass on laziness because it used to be a tech company.
The only businesses who care about the accuracy of their data are ones who depend on it to make money. In order for this to make sense, one has to deeply understand incentives.
I have a bar/restuarant where we been waiting for the verification codes for over two months now. Been trying five times to receive the verification codes for "Google My Business". No one is receiving the codes, there is no other verification codes, emails go unanswered and there is no other way of contacting a human over at Google. Our business still has outdated details and we can get no control over it.
Second case is a friend who has a store next to a market. Someone added their phone number to the market, so since about a year ago, people phone her thinking she is the market, asking for information. She receives at least one call a day and have been trying to remove her number from the market since it happened the first time. She also tried everything, but to no avail.
It's so frustrating to deal with, but there is really no alternative...
True; what makes them a monopoly is that they use their dominance in one market (web search) to make themselves "unfairly" dominant in other areas (maps).
> If there is a degradation of quality, that gives others such as Microsoft, Yelp, etc the ability to capitalize.
Except that others are handicapped by Google. Example: Average customer wants to look up a business. So, they open their browser (Chrome, 70% of market) and search, which will go to Google because that's the default (ironically, using their dominance of browsers to push their web search), and search for the business, which will then pop up a nice big box at the top showing results from Google Maps. Of course, maybe Google Maps has bad data and Yelp has better data, but since Yelp isn't a Google product it'll be at best somewhere down in the normal web results, not pushed front and center.
Monopolies are not about market share; they're about what you do with that market share.
Besides, this is so far from HN's purpose that nobody should even be considering adding this feature.