Coverage of “points of interest” in OpenStreetMap is a point of pride for many mappers. Our maps have much richer detail than commercial competitors, they provide endless handy data for mashups, and as a consequence have been a big focus of mapping party efforts in London.
But should we really be so keen? I’m not so comfortable for two reasons.
First, how up-to-date is our data? I’ve recently re-surveyed my local area in minute detail and found several takeaways, shops and banks that have closed down or changed hands. I’ve also discovered that we have very poor coverage of cycle parking in Southwark following two years of massive expansion by the council.
How likely is it that these are being regularly checked and updated? I suspect “not very likely at all”, and have therefore decided to delete all my points of interest in my local area that I’m not confident anyone will update. I mostly deleted minor shops, especially those like hairdressers that change a lot and that aren’t very important to know about. I’ve left all the amenities like banks, post offices, cycle parking and pubs.
My second concern is that the completeness and up-to-dateness will vary according to the number of active and nutty OpenStreetMappers in the area. And that tends to translate to affluent areas.
In their useful paper on the “completeness” of OpenStreetMap, Muki Haklay and Claire Ellul issue this rather stark warning:
“The large number of contributors for applications such as OSM or Google Map Maker might convey the false impression that [they] represent a real democratisation of geographical information collection, whereas the reality is that these many voices are coming from the more affluent and naturally empowered sections of society. This cacophony is likely to be silencing the voices of the marginalised and excluded even further.”
As I have auto-traced buildings in deprived parts of Southwark from Ordnance Survey StreetView tiles and sporadic re-surveying, I have noticed the very patchy and thin coverage of points of interest in those areas. Probably half the churches and schools are marked with nodes (no ways describing sites and building, though I’ve tried to draw them in) whilst the rest are missing entirely; occasionally there is a smattering of takeaways and convenience stores.
A comparison of allotments with an open dataset from the Greater London Authority reveals a similar pattern (previewed above). Most of those we have missed are in deprived areas. I’ll be revealing more work on completing our allotment coverage soon, courtesy of some help from friends and contacts in various local/regional government departments.
Muki and Claire suggest public agencies should step in to improve coverage in deprived areas, but that requires a high level of committment to OSM from those agencies. Currently we are in the very early stages on this front in the UK.
Given all of this, I would be interested to hear what other OpenStreetMap contributors and followers think. Should we bother with minor points of interest like hairdressers and takeaways until public agencies step in? Is it better to leave them out to avoid a database full of out-of-date information that only increases inequalities of coverage?
POIs are far from poisonous. Having real detail from on-the-ground surveys is part of what makes OSM so powerful. Rather than deleting your POIs maybe you could find a way to recruit more people to maintain them. Maybe the businesses them selves would be keen to maintain their details.
IMHO if you import the OS StreetView building data then it gives a phoney sense of how complete the map is, so deters ground surveys that add real value. Am I bothered that a building is there or am I more interested in what the building is or is used for? Do I want to see an over-simplified OS building outline or do I want to see a node with a name and an icon showing something I’m interested in visiting?
Hi Chris, thanks for your thoughts.
It’s a lovely idea that local businesses, community groups, etc. would maintain the data but I think it’s very unlikely, *especially* in deprived areas where fewer than 50% of households are on the internet and they have much bigger stuff to work on.
I also think it’s very unlikely that areas such as north Peckham and Walworth (home of some of London’s most infamous housing estates) will see many regular ground surveys from anyone. They’ve only been mapped with basic features after a couple of mapping parties and the efforts of a couple of mappers including myself. Given that, I think the buildings give a very tangible improvement.
Wait – what? There’s things on the ground, that are in OpenStreetMap, and you’ve deleted them? Simply because you don’t think you yourself are going to update them?
That’s bizarre. I’m not planning on re-surveying all the roads I entered in SW London, so any changes to the junctions there in the next 10 years I won’t pick up either. Should I follow your lead and delete all of “my” roads too?
The distinction I’d make is that those roads are much more likely to be kept up to date, and are pretty essential to the map being useful to anyone.
Hairdressers in SE London, on the other hand, are pretty unlikely to be updated and are only useful to people who want to see hairdressers on the map.
There are several perspectives on POIs. Here are a few: (1) POIs are among the simplest map attributes. It is very easy for (OSM) newbies to add a POI and therefore POIs can be used to attract new contributors. (2) Outdated POIs make a bad impression for first time OSM users. Therefore no POIs are better than outdated POIs. (3) POIs play a significant role in social LBS for “check-ins”. A lack of coverage of POIs means that OSM is less attractive for social LBS.
Personally I would like to see a dedicated POI editor that is very simple to use to get people familiar with editing the map.
I would also like to see something like an “official building ID” (maybe a short code of an address) to keep more detailed POI information in a separate database but having an ID to reference to it.
Your latter suggestion isn’t that dissimilar to Google Maps, which relies on its scraping to position and update POIs.
Is OSM the best place to build a comprehensive database of every last shop and takeaway?
Tom, I think you’re mistaken in putting any hope or trust in public agencies. They might step in but if they do, their data is going to be no different from that which you just chose to delete because you had doubts about its future. I am slightly surprised that nobody has mentioned what I consider the obvious solution – more thorough options for quality management, including the age of information, including, possibly, an inverse OpenStreetBugs thing where people don’t say “there’s an error there” but instead “yup, this section of the map seems right to me”. Then, depending on when something was mapped and how many people have said “looks right to me” you can compute a reliability indicator, which would then be used by the community to re-check where it is deemed necessary. Don’t worry, it’s all going to be fine even without you deleting half London.
Frederik, I don’t put much faith in public agencies. Yet. That was my point.
Your QA tool suggestion is very interesting and would certainly be useful. However, it wouldn’t address my second worry, which nobody has commented on yet: it wouldn’t be much use in deprived areas which are poorly serviced by mappers despite the heroic mapping party work of Harry Wood and could be left with a scattered selection of out-of-date POIs.
By the way, it would be nice if you didn’t accuse me of scary lies. I have only deleted a handful of nodes for very minor POIs in my local area that I originally added and that nobody else had touched. Hardly “half of London”!
Thanks for your patronising response. I guess it matches your jaded view of the people who live in ‘deprived’ areas.
My jaded view? Odd. I’ve just followed the analysis by Muki & Claire, and my own person observations having put huge amounts of effort into mapping the deprived areas around my local area of Peckham, and various efforts to engage with the council, tenants’ and local resident groups I’m involved with.
We can generalise this level-of-detail problem, and in fact this repeatedly comes up in discussion in different guises. POIs sit somewhere on sliding scale of detail, where at one end we have essential map features, and at the other end “map every blade of grass”. Mapping trees, park benches and lamp posts? They seem pretty nutty to me. It’s not just node features. Do we map pavements? We could connect them at the highway=crossing, to help with pedestrian routing. And if they’re very wide pavements, do we map them as areas? And if we’re doing that, why aren’t we mapping roads as areas?
I think it is maybe a problem with OpenStreetMap, that there’s not cap on level of details, so we can seek to add more and more, and never really drop down a gear into data maintainence mode. Collectively we’ve almost managed to agree not to map pavements… almost… I have seen people mapping them in places.
I’ve always encouraged mappers to “Do the kind of mapping which you find fun”. But perhaps we should also encourage mappers to think carefully about what level of detail is consistent, or attainable across a larger surrounding area. To me address mapping feels like a rather unattainable level of detail. It feels like a shame that people waste time on it. Building outlines are bit crazy too, but we can at least aim for consistent coverage, by plugging the city centre gaps. POIs are more tricky, because they’re not evenly distributed.
As far as updating goes, it is a different kind of contribution. I personally find it quite fun when I spot a POI which needs updating, but the idea of an expedition purely to check existing data seems boring. With enough people joining the project, and with the tools getting easier, hopefully these things become more feasible. One OSMer’s big mapping area today, is the local neighbourhood for several OSMers tomorrow. A lot of people edit wikipedia to correct typos. We need to get the same dynamic in OpenStreetMap. If we can make that happen, then we’ll get more activity in less affluent areas. But I think there’s a danger that we’ll always be destracted by mapping the next detail level.
Thanks Harry, some very interesting comments.
I have been doing a lot of building and address data of late, *slap on wrist*. Actually my reasoning is that I find it fun (!), I find it useful, and I assume that those things won’t change very often so it’s very low maintenance.
I’ve been thinking about trees, I know you think it’s absurd overkill but again I personally think it would be useful, interesting and fun. Not sure about keeping it up to date, mind you.
It would be very interesting to think how a mapping party focused on maintenance could work. Or how we could engender a greater use of the QA tools that are around now that most of London has been covered to make detail more consistent and accurate.
I think you’re right that it’s worth thinking about attainable levels of detail that can be consistent, especially following the analysis of Muki and Claire. I’m quite disappointed in many of the comments on this post, which suggest that almost nobody else is willing to contemplate that. Perhaps it’s a difference between seeing OSM as a personal hobby of many individuals, and seeing it as a collective effort to provide excellent, consistent map data?
POIs, in particular, shouldn’t be all that much of a problem for maintenance. Why? Because everyone is able edit them, and there are many people who will regularly come across these POI and notice outdated data (unlike, say, some details of remote hiking trail).
Correcting an outdated POI is an easy enough task for a beginner or a casual “typo corrector” kind of contributor. You don’t even need a full-featured editor for this. Polish the Amenity Editor (http://ae.osmsurround.org) a bit and even the new shop owner themselves will be able to fix the data.
Of course this requires an adequate level of exposure for OSM so enough eyeballs will check the map for errors. But that’s something we would like anyway, isn’t it?
That being said, I usually do not map minor points of interest, because – to use Harry’s words – that’s not the kind of mapping I find fun. I’d rather map pavements, for example. 🙂
The thought “seeing a problem when there isn’t one” comes to mind here. No map database will ever be 100% up to date and accurate because change happens all the time. A low, insignificant percentage of error does not really matter to most people. Finally it is the USER of the data who will ultimately decides whether it is useful or not, NOT the data PROVIDER
This has a strong resemblance to the “relevance criteria” or “Notability” of Wikipedia ( http://en.wikipedia.org/wiki/Wikipedia:Notability ). A topic that has been rather hotly and controversially debated and has caused quite a bit of negative attitude and publicity towards wikipedia. Although I suspect eventually OSM won’t get around something similar once every tree, every cobblestone and every grass blade gets mapped, I really hope that geospatial position of POIs such as shops don’t overwhelm OSMs capacity to keep things up-to-date and thus really shouldn’t be deleted.
However, yes, I do think that OSM in general, and London in particular, increasingly has to move over to “maintenance mode” rather than (re)surveying. So far I don’t think we have done a particularly good job at that, and even supposedly well mapped areas such as London still contain a significant number of errors, not least, because tagging schema have evolved and expanded over the last 5 years.
We need to find better tools, ways and processes to notice bugs and make it easier to fix. But I think particularly we need many more contributors. Whereas an initial survey of an area can be done with a small number of people, data maintenance (presumably as it is less fun) imho requires much larger numbers of mappers to achieve. At least in London, we don’t seem to have reached sufficient numbers yet to keep things up-to-date and bug free yet.
So rather than going and deleting POIs and thus setting the “Notability cirteria” as high as that, we need to improve tools and convince many more people to use and contribute to OSM.
Your deletion should be reverted.
I find it very valuable to know the exact location of these shops, and one of the great advantages of OSM.
I think it is currently a bit of a chicken and egg problem. With more people entering this kind of data, you’ll slowly get more people using the data, which in turn means you’ll get more people taking an interest and trying to update the data and make sure the data is up to date, with simpler editors coming along too. Mapzen POI Collector is one such editor that is making it easier to do this kind of POI surveying.
I predict that this kind of data will take another year or so before the critical mass of people using and updating the POI data, however without it being there, people aren’t going to map it and potentially update it.
It’s not the first time I’ve heard someone say “Oh I didn’t realise that you could map that” after seeing something that someone else has mapped. Thus without these example areas of detail, you won’t start getting people creating more of them and then starting to look after their own area.
If you fancy reverting it, here is the changeset, although a few of them have actually closed.
http://www.openstreetmap.org/browse/changeset/5550488
Hi Tom,
This is an interesting discussion.
On the specific point about the paper – what we tried to point out is that there is an interesting potential that is emerging from our analysis of OSM and other similar data sources. What we are seeing is that popular and affluent areas get attention by the mappers. They collect and update them regularly. Deprived areas, for a range of reasons, are not well covered. therefore, for map updates, public sector bodies can trust the crowd to provide them updates about changes in central, affluent areas and put their own limited resources into updating deprived ones. We were not suggesting that the public sector will use OSM or update it. On the contrary, the analysis points exactly to the problem that any public sector body, who, by their nature, need to deliver more services to deprived areas, will face if they are going to rely on crowdsourced information. Probably this point in the paper requires more clarification …
Now to the issue of POI – I would argue that you are not improving the situation by deleting POIs. At least from the point of view of inequality, it is important to show it and not to try and ignore it or hide it. It is much more valuable to put an attention on the places that are not mapped and instead of mapping them yourself find the local people who will map them. There is a lot of value in using mapping as a way of encouraging people to notice their area and to change it. I would even guess that it is possible to raise some funding for such effort.
This is one of the facets of OSM that shows that in terms of motivation it is more of leisure activity and there is a certain lack of real sense of civic commitment – if Mikel managed to raise the funding to map Kibera, there shouldn’t be any justification not to map any part of London…
If you don’t want to see shops that are possibly out of date, don’t render shop=*.
With easy to use tools, and potential new tools, it can become a lot easy to select (perhaps from a check list rather than the usual!) a shop and say it has changed it’s name/function. You can do that in one step, just by knowing what shops it is next to or how many shops down the road. One can also be surprised and how many people want to get into OSM-editing because they know something is wrong/outdated (or they nag at me to change it). Keeping the nodes/areas in also keeps the history, what some of us are interested to see could be an exciting resource (in 10 years time, can we see how London used to be?).
Someone joined OSM in Durham this month. They wanted to start a business so were surveying similar shops in the city centre and decided they would contribute to OSM at the same time. For some odd reason I haven’t made a good shop pass, but I really wish I had. I wish I had added as many shops as possible, and with building outlines for them. It would make it easier for them without a GPS, they could also update ones that had changed. Even if they considered it could be out of date, they could run a query to count how many shops were in the area.
I am currently adding buildings and house numbers/shops across Durham. It’s a good way to see it what has changed in <=3 years, or what I missed (or mapped badly). I've found this exciting enough that in 3 years I will probably go for a stroll/cycle through places I wouldn't normally, just to see if they've changed (probably not a full sweep, but one to get an indication).
A maintenance mapping party would be interesting to see indeed. What area of London have we not been to for the longest time? How many edits can we make rather than new nodes/ways/areas?
In Romania i try to add as much POI as i possibly can.
Why ? So that people that use there mobile phone install an app that uses OSM data and they try to find the nearest X (hairdresser, furniture store, etc ) to find.
Also for GPS devices.