It’s the start of a new Quarter, and so it’s time for a new UK Quarterly project. The project for 2026 Q2 is Addresses.
For ideas, suggestions, datasets, tools and resources, have a look at the wiki page:
https://wiki.openstreetmap.org/wiki/UK_Quarterly_Project/2026/2026_Q2_Project:_Addresses.
Feel free to contribute to the page and also discuss things in this thread.
We’ve got a lot of buildings which now have addr:postcode and ref:GB:uprn, where adding addr:street should be quite straightforward.
Each postcode should relate to the same street, but there are occasionally some addresses which also need addr:substreet or addr:parentstreet.
Is this something for which a MapRoulette challenge could be created?
Please bear in mind that there are plenty of cases (typically in older settlement centres) where a postcode spans multiple streets, or only addresses a ‘street’ which is only accessible by foot rather than vehicle (so may be mapped as a footpath). As you say, there are also parentstreet situations.
The UPRN dataset seems to use the nearest postcode centroid for UPRNs which don’t have an explicit postcode associated with them, so blindly copying postcodes from the UPRN dataset is not advisable without verifying them some other way.
In particular, the UPRN dataset tries to assign a postcode to everything with a UPRN, even things like bus shelters or ponds (which obviously can’t receive post). Please can people not add postal addresses to these items! postal_code is appropriate for some of them, but it’s probably better left off entirely.
(I’m sure you know this from your work on UPRNs, but others likely will not.)
As @gurglypipe says, there are some important edge cases that need to be considered.
My understanding is that the postcode for a UPRN is the exact postcode (provided it is also in CodePoint Open), or the closest postcode if not. I am not clear if postcodes that are not in CodePoint Open are empty or the nearest in the UPRN dataset, but the documentation suggests they would be empty. For most buildings this should be correct, but for non-addressable buildings the value may be somewhat arbitrary. I avoid adding addr:postcode to these, but adding postal_code may be appropriate.
Where the same postcode is linked to multiple UPRNs this can often be a good indication that a street has the same name. This can be handy to determine where a street changes name along the way, or where errors have been made in previous mapping.
In areas with very few houses, a postcode may cover multiple streets. You can typically see this at small hamlets on a crossroads, but other cases do exist.
I’ve not yet found a case where a postcode in a reasonably densely populated region of housing does not exactly relate to the street, but I would not be surprised if there were some. I’d be quite interested in any examples of this, to have a closer look if it can be inferred from the data.
My understanding is that the postcode allocation to UPRNs in the ONS UPRN Open Data products is as follows:
This means that any UPRNs that have postcodes and correspond to a current addressable property should have the correct postcode.
UPRNs that don’t have a postcode assigned could be new addressable properties with a new postcode that doesn’t yet appear in Code-Point Open. Or (I think) they could be old addresses where the postcode no longer appears in Code-Point Open.
UPRNs that correspond to streets or non addressable properties/objects will have a nearby postcode assigned. Since the object doesn’t have an address, these postcodes aren’t really correct or incorrect in any meaningful way.
Hello Mr Robert Whittaker,
Are there any specific Hashtags for changesets for this project?
@0235 I don’t think anyone has set any yet. Would you like to propose one?
(I don’t know if using hashtags on a project like this is useful or not. Even if we agree on something, I’m sure lots of people won’t bother to use it, and does anyone actually use the information for anything?)
If there isn’t one set, then may not be worth it. I think from a few years ago there was a “#OSMUK2-26” or something like that for defibrilators.
I’m quite new to OSM and I want to help add buildings, their housenumbers, and streets to the map as part of the challenge because it would help make finding places easier.
What is the benefit from adding UPRNs or postcodes to buildings (if they clearly match location and are from the appropriate source of course)? I’d rather spend time adding numbers and streets if that’s more useful. Similarly since I don’t want to knock doors for postcodes, is there any point me adding any from the centroid data? There sounds like there’s loads of edge cases, and anyway I’d only be doing the estimation that Nominatim does. Basically, which address data is most helpful.
Thanks!
Yes, there is a point adding postcodes from CodePoint Open, although you may not be able to trust it much further than the building nearest to the centroid.
Both are useful. Where we have the UPRN and postcode, we still need to get the house number and street from a survey or (for some older buildings) from the NLS OS map layers. They’re not exclusive, I do both.
The datasets OS Open UPRN and ONS UPRN Directory are available under the Open Government Licence, so can be used in OSM. These (mostly) allow us to add the UPRN and postcode for each property. So there’s no need to survey postcodes at all. (The centroid data in OS’s Code-Point Open, is largely redundant.) I’ve created a convenient viewer for this UPRN and postcode data at GB Unique Property Reference Numbers (UPRN) | Rob's OSM Stuff .
But as with everything in OSM, please focus on adding whatever information you are motivated to add. In terms of addresses, the most important things (that usually need a survey, and hence take the most effort) are house-number/name of each property. The street name is probably the next most useful. (This can often be determined from the above datasets and existing mapping, but is often a manual process.)
Postcodes are certainly useful to have, in terms of returning complete addresses for data users. UPRNs less so, from a data-user point of view. But they can be useful in checking for completeness of mapping and verifying postcodes.