Saturday 4 March 2017

Postcode changes

I have provided a tile layer of postcode centroids based on the Codepoint Open postcodes since they were first published in 2010. I also provide a similar layer for postcodes based on the Office of National Statistics dataset called ONSPD. I have just updated the two layers with the recently updated data.

The ONSPD dataset contains almost a million more records than the Codepoint Open one. Most of these extra codes are retired postcodes. There are also some entries for BT codes. These are Northern Irish postcodes which are not released under OGL, so I don't publish them - I wouldn't want them to be used as a source for Northern Irish postcodes in OSM as that would violate their copyright. There are some IM codes, but without any coordinates. These are Isle of Man codes. I don't believe they are released under an open licence so again I don't publish them, but having no coordinates makes them useless for this purpose anyway.

The ONSPD dataset has a column for the start date and one for the end date. The earliest start date (just year and month) seems to be 1996-06 which most active postcodes have, the most recent start date is 2017-01.

Edit: The spread of dates is much wider than I first thought, with the earliest is 1973-08.

I think showing any newly added postcodes would be useful to help people see where there may be new developments that need surveying. Some new postcodes will just be yet another code on a building that already has many postcodes. These are places that get lots of post (after all that's what postcodes are for), often royal mail buildings where there are PO Box addresses.

The recently published files have changed their formats (again). Now the Codepoint Open dataset has fixed length postcode fields, so the natural spacing is lost unless they are reformatted, which is very easy. In the past the list of postcodes from Codepoint Open exactly matched the processed list of postcodes from ONSPD, now they don't. I'm examining what it is that is different.

Many of the centroids have moved a few metres. I can understand that a centroid might move if the street or section of street had some new houses (delivery points) added or removed, but that is not the cause of most of the changes. It seems the reason for the change is that the source of the centroids has changed. The data used to come from Address Point (which is now discontinued), now it comes from the Postal Address Location Feed of Geoplace. This is interesting because it removes the connection with Royal Mail and gets the postcode from local authority data. Local authorities are where addresses are created, though Royal Mail must have a part to play in adding the postcode. Getting the data from Geoplace may be one small step towards releasing UK address data as open data. I hope so.

To see how to use the tile layers follow the links for Codepoint Open and ONSPD.

I have discovered there are 111 postcodes in the Codepoint Open dataset that are not in the ONSPD dataset after I have processed it. These are 'large user' postcodes and they don't have useful coordinates. In the Codepoint Open dataset there is a field called positional quality indicator. 10 indicates the best positional quality, 90 indicates the worst. All 111 are 90. In future I'll ignore these entries.

No comments: