This month we launched our long-awaited, highly-anticipated UK Places data. While we’ve been curating points of interest (POI) data in the US for the past few years, our customers have been asking for the same in the UK. Our UK data offering provides our Core Places and Geometry datasets for England, Scotland, and Wales, covering over 1.3M places and more than 500 brands.
We have big aspirations here at SafeGraph. We aim to one day provide places data for the entire world. Our expansion into Great Britain taught us a lot about how places - and as a result, places data - differ by geography. In particular, we encountered five main technical differences between US and UK POIs.
American jargon does not always translate to British jargon very well, and our machine learning models required some lessons in the Queen’s English to make use of the metadata attached to UK data sources. For example, humans may know that pubs and inns in Great Britain are the same as bars in the US, or high streets are the same as shopping strips, but computers need to learn these things.
Postal codes across the pond are much more granular than their counterparts in the US - so much so that some high-rises comprise several postal codes. Our US POIs represent 37K distinct postal codes, while our UK POIs represent a whopping 609K.
Great Britain is crowded! POIs are more densely co-located in the UK than in the US. Great Britain has 14.9 POIs per square mile while the US only has 1.7 POIs per square mile. Of course, the US has a much larger area than Great Britain, but the resulting increase in POI density was a new challenge for us as we built out the UK Places data.
With these crowded, co-located POIs come smaller spaces. The average polygon size for branded, “OWNED_POLYGON” POIs in Great Britain is 934 sq. meters. It’s roughly double that in the US, at 1,917 sq. meters.
Great Britain’s architecture is much older than America’s (some buildings date back to 3000 BCE!), making distinct buildings much harder to delineate. We still strive for world-class polygons that define even the most obscure demarcations between adjacent buildings, but it’s a much taller order, and there will inevitably be more “SHARED_POLYGONS” in the UK data as a result. Users can expect our “% OWNED” polygon metric to increase over time as we continue to learn more about this.
Building places data in the UK has been a fascinating lesson in geography, history, and culture, and we are excited to see what other regional differences we encounter as we continue to expand.