Hacker News Reader

I’m an ex-Googler and I tried to fix cycling routing. I’m honestly sorry I wasn’t successful. We had a really nice solution but privacy considerations made it extremely difficult and limited. And then I and the other main driver both left Google so it withered and died I believe. I really hope someone else can fix it one day.

This is a throwaway obviously since this basically 50% doxes me.

> We had a really nice solution but privacy considerations made it extremely difficult and limited.

Would you be willing to elaborate a bit more on those privacy considerations?

(I guess what I'm struggling with is how cycling is different in this regard from other modalities.)

There are significantly more people using the other modes, especially in active navigation. For example, it’s well known that Google detects traffic through the movement of active users (there was the wagon of phones traffic jam experiment). Imagine trying to do something similar with the amount of cyclists actively using Google with the necessary privacy settings enabled on any given route. It would be noisy but possible, but then on top of that Google is extremely strict with privacy requirements (the exact opposite of what everyone on HN assumes). And once you add all of Google’s voluntary restrictions like anonymity of inferred data and such, it’s nearly impossible to gather any signal, even in major cities.

Edit: By the way, the massive difference in potential users compared to all other modes is also why it’s so hard to prioritise work on cycling inside Google.

This sounds a bit American to me honestly - I would think that in Europe (and especially in Amsterdam) there would be enough cyclists to make the data usable. OTOH, they may not have the necessary privacy settings enabled (I think that's a minority among car users too). Of course it's also a vicious circle: if Google Maps is bad for cyclists, few cyclists will be using it, so you won't get the data you need to make it better. Additionally, I think the proportion of cyclists using any kind of navigation solution is lower than with car drivers...

How many cyclists use Google Maps to navigate? I guess (with no true facts) most cyclists know their primary routes and when going somewhere new have a rough idea and just check for the last part.

Whereas car drivers often use navigation all the time to see traffic and alternatives, which are non-issues for cyclists.

Thus even with many cyclists the information they collect is less.

Exactly.

Add that there’s no Android or Google Auto for bikes, you usually don’t charge while biking, etc. The people who do need to look up their cycling route typically do it beforehand, so in the end even in cycling-heavy regions there are very few using active navigation.

Delivery workers will in regions where that’s common, but they likely use company-provided or other local solutions.

Biking through Keskuspuisto central park, even when born and raised and lived in Helsinki for my whole life, one still needs to consult the satellite map a lot. It's a maze. The signs are roughly indicative at best. So many places look exactly the same you forget the exact crossings during the winter. You often can't see far because it's a forest.

> How many cyclists use Google Maps to navigate?

The Cowboy bike company have started using Google Maps for navigation in their app, which is a bummer: Before the change, the navigation would reliably take direct me to use roads that a bike friendly. Now it does just the opposite: it wants me to take big roads that are only friendly to cars.

Google Maps was a nightmare at the top 100 US News university I went to. It'd constantly be unaware of bike routes and try to take me an absurdly wrong way around. Even as I was taking the bikable shortcut paths it would just keep rerouting telling me to turn around 180° even when I was more than halfway near the other road. You'd think there's be sufficient data on a major university campus...

> and just check for the last part.

hit is not how I use navigation apps when cycling. When I don't know the way, I start my navigation app (not gmaps) when I drive off, because

- I don't want to stop midway to check where I need to go, when I can do that when I get on the bike - I don't know where exactly to stop to check where to go - If I don't know where exactly I want to go, going roughly in the right direction probably means that I cycle longer than I need to

Also mistakes driving a car are more costly. If you miss a turn on a bike, you can stop, walk your bike back, and take it. In a car, if the street is busy, you can’t do that, and who knows where you end up if you take the next exit etc.

You think Google is watching your location only when you're using Google Maps?

The widespread assumption that Google collects all data all the time and uses it for everything is ridiculous.

There’s a labyrinth of explicit user agreements, all of which are strictly enforced, and then there are further layers of voluntary restrictions on top of that.

Yes, Google might collect location data even when you’re not using Maps (there are a lot of passionate, dedicated Timeline users), but you have to have that enabled, and even then the data can only be used in certain ways.

I had to go to great lengths to disable location history, and Google penalizes me for it by disabling search history in Maps as a result. I'm sure 99% of people don't go to great lengths to disable location history and so yes your entire history of GPS locations is basically saved to a Google database as far as I remember. (Note I'm obviously a user recalling details from years ago and not a Google employee/developer. My perception that Google collects all our GPS details forever by default is based on reality, whether or not that means you as an employee would actually be able to do anything with that data.)

> The widespread assumption that Google collects all data all the time and uses it for everything is ridiculous.

Maybe they fail to use it competently but it definitely collects a lot of data and attempting to stop it results in punitive reactions and disabling of features not needing it - and repeated badgering to permit Google to collect data again.

Google also collects data that I was initially unaware that it is collecting.

"Yes, Google might collect location data even when you’re not using Maps (there are a lot of passionate, dedicated Timeline users), but you have to have that enabled"

Of course that opt-in requirement was only the result of a hard-fought class action settlement.

Take this with a grain of salt as my recollection of the litigation is somewhat hazy. But I believe Google argued, even when a user opted out of location tracking, Google could still keep the data for its own uses. Ie opting out of location tracking simply meant the user didn't have access to their location history, it didn't show up on the phone (couldn't be searched by suspicious spouses) etc.

I believe they also collect location data when you're not using Google Maps for the purpose of updating their database of wifi locations (that they then use for location tracking). There is an agreement for this too, but the Android phones I've had regularly spam you until you agree to it.

You mean: until you accidentally click to agree because it popped up just when you wanted to click something below it. And then of course the only way to disable it again is hidden somewhere several levels deep, and comes with warnings that it will break your phone.

Well, they collect quite some information, but to figure out which exact routs are taken etc need a persistent collection which is relatively expensive. The "passive" cell based location tracking doesn't work for that.

In a car you can charge your phone and battery usage for GPS isn't as much of a problem.

The ex-Googler is most likely talking about the product as a whole, across the globe. It would be hard to justify features dedicated to a small group of users overall (with different goals/expectations), especially while keeping privacy (and data sharing). That being said, Google maps works great for cyclists in the US ...

It's maddening that tens of millions of users is "give up and do something else" territory for Google.

Yes, it’s also maddening for most Googlers who are passionate about the work they want to do that would affect just tens of millions (or less).

But at least it leaves room for other companies in some situations. What’s really sad is when there’s something that could only be done by Apple or Google, but it’s too small for either of them to actually do it.

Problem is those companies are like Nest, who made good products with love and then sold themselves to Google who then turned most Nest products into bargain-basement offerings with half the features. So the only real fix is FOSS, since everything else can be bought and killed. (FOSS can be too, but it's harder.)

Supporting sustainability, also a job for someone else. We are so lucky OSM exist, helping to perhaps save the cities and the planet.

> Additionally, I think the proportion of cyclists using any kind of navigation solution is lower than with car drivers...

In large part because cycling is fairly resistant to traffic jams

I don't think it's necessary to jump right to data collection, the routing algorithms and incomplete maps are the issue. Of course, you'd need additional data about bike lanes, etc (which could come from crowdsourced data? Maybe that's what you are alluding to?).

As it is, offline routers such as Brouter and OSMAnd do a much better job, and it's pretty easy to convince other cyclists to use them.

Traffic jams are almost never an issue when biking - heavy car or bike traffic can slow you down, but by that point you have at least the same density as car traffic jams for data collection...

Is it that there are more cars? Because I live in Amsterdam, and there's definitely more bikes than cars here, yet bicycle support still sucks. Although it's possible few cyclists use Google Maps because a) it's bad, b) they don't need to, and c) you need a special clip to use your smartphone live on your bike (though that's also true for cars).

Most modern cars have a center console with Android Auto, so you just plug in your phone to USB and it takes over navigation, communication and entertainment.

I guess my car isn't that modern. Our car nav is okayish, but the ui is crap and the maps are outdated.

What exactly makes the challenge of cycle so much harder than walking? Also is there really 0 ways you could use pedestrian mode data for biking directions? Even if marginally or to confirm hypotheses mostly based on cycling data

Lastly, is there anything stopping you from contributing potential innovations to open source alternatives?

In most densely populated areas near where I live in southern New England, bikes are vehicles that, in most cases, must operate on the roadway. I'd get a ticket for riding a bike across a pedestrian bridge or down an urban sidewalk.

Two days ago, thanks to cycle routing, I accidentally discovered a bike path (with some dirt even!) that ran parallel to a more busy car road that I have been riding for 35 years now. I had no idea it existed. Not all seems lost.

Also, in Amsterdam it works pretty well if you set Google Maps to "walking" and then divide all time estimates by 3.

Google maps is still terrible for pedestrian / cycling use.

OSM is much, much better but I guess Google refuses to pull OSM data because it would require them to open up their own maps.

For cycling, yeah, that’s why I said I’m very sorry I wasn’t successful.

For pedestrian, I’d disagree. At least in places I go, I find Google’s walking directions to be top notch.

For hiking OSM blows Google out of the water. I'm surprised how detailed OSM is with random forest trails and shortcuts, resting spots, viewpoints, etc.

The main thing I use Google Maps for is searching for stores etc in cities. I hope such information becomes easier to contribute to OSM.

Perhaps a Pokémon Go style game for map updates could work? StreetComplete[0] is good, but the rewards are kind of boring.

0 https://streetcomplete.app/

Recall that Google as a company doesn't care about making people's lives easier and better, they care about inserting themselves as a middleman in aspects of your daily life and then trying to sell ads about it. They might invent a "talk to your grandma anytime" app and tell you it's because they love your sweet grandma, but the only things that will get attention and dev work are the bits that figure out how to sell you more stuff.

If you're a pedestrian in a city and you're walking between two locations with mailing addresses, then yes, Google is great. For anything outside of those boundaries, OSM is head and shoulders above Google Maps. GMaps is basically unusable for hikers and outdoorsmen because of the asinine offline UX and lack of topo data. I've also seen it make some outright unsafe routes when walking around suburbs and smaller towns.

Google fails pedestrians for the same reason it fails cyclists - it’s off road data is terrible & it’s not interested in fixing it.

In large cities, the roads are the only routes, so pedestrians can go OK with GMaps. Anywhere else & Google is prone to send you waltzing down a main road instead of the paths marked for actual pedestrian use in my experience. Same for cycling.

Thanks for trying, at least: You weren’t able to shift leviathan, but there’s no shame in that.

I’ve been wondering why doesn’t google use its own elevation and traffic data? There are no privacy issues with the data they already make publicly available. Prioritizing flatter and more pleasant routes over shorter ones couldn’t be more complicated than adjusting a few variables, could it?

Google displays my street name wrong and there is no way to change it. I've reported it 10+ times and nothing is happening. I don't know where they got that street name from but it's definitely not the "official name" of the street as recorded by the official authorities. And this is in Europe on a street where the google car goes through basically every few years.

I can't even describe the level of frustration this causes because I'm forced to use a non-existant street name for packages because otherwise no courier finds me.

But with Organic Maps, since it uses OpenStreetMap, I've actually corrected a bunch of street names near me and have added about 4,000 house numbers. It's really nice when you're able to improve the area near you.

You're doing the Lord's work.

That is really harmful and bizarre. But that also means you basically need to know which delivery service is going to deliver it, because some, like the postal service, may be relying on the real name or the postal code.

What's the street?

The edit review system for Maps is really frustrating. Sometimes obscure things are approved instantly, and sometimes really important things are silently ignored.

I once attached a news article to my report, and they still denied it. I've also added street vendors like "coconut cakes lady" and had them approved.

Make sure it’s correct on OpenStreetMaps. I believe Google it as one of their data sources, and it’s something you can actually go in and fix yourself.

I recently moved to Amsterdam and I find Google maps biking directions are often wrong or just a bit worse than organic maps. But also when organic maps it wrong (like one time it though there was a nice connecting sidewalk that no longer existed) it was super easy to go edit on OSM and next time I'll get the right directions.

My one wish was the OSM had a great app for discovering POI or reviews. Like something that can show me all the bookstores around but with a little extra human touch on if it's nice to sit at or mostly used books or something like this.

Here you go: https://mapcomplete.org/shops.html?z=12.2&lat=52.36511&lon=4...

Note: it takes a little while to load all of Amsterdam.

Disclaimer: I'm the (main) dev of mapcomplete

When did you last refresh your data? It seems at least a year out of date compared to OSM as there are shops I've added that aren't showing.

Confirming the outdated map data lacking many changes that are present in OSM.

I feel like nobody in the Google Maps uses a bike for commuting or other daily life purposes. If they did, they wouldn't route me through the most dangerous, fast traffic street in my neighborhood, and instead they would choose the adjacent bike boulevard. Apple Maps does this much better, even though they've sent me to the wrong side of town a few times.

From my early tests, Organic Maps also sends you down dangerous routes, so I'll use cautiously.

The bike profile in OSM routers will prioritize bike paths, bike lanes, and small streets over big highways, but there's always a balance since you're unlikely to want to bike 10 miles out of your way to avoid a busy street. That balance is always in flux and can only ever be as good as the OSM data underlying it (road classification and bike tagging etc)

If you bookmark streets that it's sending you down or avoiding inaccurately and check them out later on a computer, I bet you'll find that there's some OSM tagging issue. https://osm.org has a routing tool you can compare with as well. The OM Telegram channel is always willing to take reports on where OM routes are worse than OSM.org routes, and OSM channels are happy to help figure out what tagging issues there may be.

Cheers!

The balance is problematic in every cycling directions app. Some may be a bit more comprehensive than others in certain areas but all of them prioritize shorter routes over flatter ones.

I wish there were a true flattest route app/website that prioritized quieter roads and more gentle slopes by default using existing terrain data instead of making users manually add a bunch of waypoints using their own knowledge in order to compare elevation profiles.

I know that OsmAnd has fairly granular routing preferences, including a "preferred terrain" setting with "hilly", "less hilly", and "flat" options (as well as a "use elevation data" option which I don't quite understand).

Magic Earth (proprietary app that uses OSM maps) allows you to choose the maximum hilliness you want your cycle route. I've never used cycle routing, so can't comment on how well it works.

I they they have designed a routing algorithm that works very well in cities where Google employees live and work, and they either haven't tested elsewhere or aren't willing to make targeted changes to fix specific areas. It does work very well in US cities that have Google offices, but it's not surprising to me that it would fail elsewhere.

I mean, you'd think that, but lately it constantly tries to send me down illegal lefts, and tries to have me drive on the Muni/Taxi only part of Market. In San Francisco. And if San Francisco isn't a coty where Google employees live and work, I don't know what is!

This seems local. I just checked routing from North Berkeley to Emeryville and Apple Maps routes me down the main north-south traffic sewer instead of good parallel bike routes, and Apple Maps appears unaware of the new bridge across the rail yard. Google's results are far more sensible.

Exactly this. I'm sent down Sacramento or San Pablo regularly, instead of the California or King st boulevards. Most people who ride around don't need the directions, but now and then you see someone you can tell is following them, not realizing there's a far safer option around the block.

The amount of mistakes Google makes in the Netherlands when it comes to bikes is mind-blowing. If you don't know the area or manually check the maps it takes you trough a lot more car places that it should do.

I've had some horrendous experiences using Google Maps for cycling in the UK too. Mostly it's very good, and it's impressive how good it is, really, but it's the times it gets it wrong that you remember.

Routing me through a park in the middle of the night, when it was closed, was one. Taking me on a detour to avoid a busy road, which turned to be on a farm track where the road surface was completely destroyed, was another.

Yeah, I think I've gotten too used to it and learned to automatically compensate for it, but the more I think about it, the more I realise just how terrible Google is at bike routes.

How does https://cycle.travel/map/ perform routing-wise in Amsterdam?

(Unaffiliated happy user.)

Maybe slightly better, but still not great. It's skipping the eastern part of the Sarphatistraat. Also it struggles with street names and encourages you to only type a town, which isn't very useful.

I like that it allows you to choose between paved, gravel or any. Although gravel may still be too limited for some; mountainbikers like dirt tracks, and I've had Google Maps send me through loose sand that would probably require a BMX.

> Also it struggles with street names and encourages you to only type a town, which isn't very useful.

Oh, good spot. Fixed. OSM geocoding is better than it used to be - I used to just say "town" because streetnames were partial, but they're pretty good now.

If you have a start and endpoint that should use Sarphatistraat but don't, I'm all ears. (It's my site!)

That's an amazingly quick fix!

Try "Feike de Boerlaan" to "Vondelpark". I would expect the trip to take the Czaar Peterstraat and then the Sarphatistraat, but instead it takes the Panamalaan, Zeeburgerdijk, Mauritskade, and only switches to the Sarphatistraat at the Weesperplein, which I think is also the point where the Sarphatistraat changes from a bike street to a regular street with a bike lane. Although it later switches back to bike street again yet the route stays on the bike street.

So I have no idea what's going on with that first part of the Sarphatistraat. It's a really nice route, yet all apps hate it.

Really interesting test case - thanks!

cycle.travel's scoring for the two routes differs by 0.4% so it's very close. In OSM data terms, the challenge is that there's no particular special tagging for the Sarphatistraat - it just gets classed as a "cycleway", same as the route along Mauritskade. Plus there's a couple of very slight advantages for the Panamalaan (which is tagged as part of a designated longer-distance bike route) and the Zeeburgerdijk/Mauritskade (which are beside water so get a scenic uplift).

I might look at doing some Amsterdam-specific routing tweaks: I doubt the longer-distance bike route consideration should be relevant in this case, for example.

It's true that Mauritskade is besides water, but it's also next to pretty heavy car traffic. Is it possible to give it a downgrade for that?

> I doubt the longer-distance bike route consideration should be relevant in this case

Or maybe it is, but then it's relevant for all routes involved, including the Sarphatistraat (which is short distance for cars, but long distance for bikes).

Also worth pointing out: the very first part of the Sarphatistraat that's skipped is a bike path between a grassy tramline and water, with the zoo on the other side of that water. Can't get much more scenic than that. It's significantly greener than the corresponding bit of Mauritskade.

These would also be relevant in NYC - sounds like 34th Avenue meets this criteria

Useful test case. I've just tried it in mapy.cz which is my prefered mapping app for cycling directions when I'm out and about and it too makes the same mistake: https://en.mapy.cz/zakladni?planovani-trasy&rc=9LCtmxhwst9Ks...

Worth trying out mapy.cz app for cycling. It uses OSM data; and I find it works well in the UK, I've never tried it in the Netherlands though. It has good route finding, but unfortunatelly you can't load in a GPX route: https://en.mapy.cz/zakladni?mobilepromo=1&x=15.6252330&y=49....

there is import GPX feature when you sign in, I use it quite often.

Ah thanks; you can load them through the website and sync from the app.

OSMAnd + Brouter is really the gold standard for bike routing for me, though the Brouter Android app could certainly use some polish (it barely looks like an Android 1.0 app).

If you just want to try the routing, there's brouter-web too ;)

One feature I definitely enjoy is the ability to connect with my OSM account and easily submit changes on the go (business hours, etc).