Two startups ago (circa 2005) I created Urban Mapping. Its initial product was a database of spatially-referenced boundaries that define informal space. Since the utility of postal codes outside of delivering mail is near zero, neighborhoods are a smart way to organize and search when local matters.
While I sold the company in 2015, some of the lessons of neighborhoods seem to have gotten lost, so here’s to making better product in 2019! The talks from the first O’Reilly geo-nerd fest in 2006 are still pretty damn relevant. Standards evolve, but data isn’t much better and advertising models are still selling us a pipe dream.
Easy to say in 2018, but long ago this was far from obvious. When entering “coffee in Tribeca” into Google around 2005, you’d likely be greeted with a page of results that have everything to do with the eponymous Noguci coffee table and nothing to do with a cute cafe.
This is because five digit strings, aka ZIPs, have been geographically disambiguated to points/bounding boxes/polygons for a map, but neighborhoods remained unrecognized in an explicit geographic sense. Urban Mapping’s first product was comprehensive database of all socially-defined space, starting with areas of high density.
Prior to bemoaning what is correct/wrong (difficult as rarely is there an administrative/political definition), it’s instructive to understand how people think about informal space–why all the consternation, community uproar and divisiveness? Over at Bostonography there’s an interesting project on collaborative definitions of informal space (though they don’t call it this), ie, neighborhood boundaries.
It’s somewhat analogous to another project that also sought to define crowd-sourced neighborhoods. The esteemed Dan Montello (and colleagues) of UCSB published a paper about dealing with “fuzzy” and vaguely-defined space, which provides a great intellectual foundation for understanding the issue. Of course Kevin Lynch’s The Image of the City was the first modern work to capture elements of urban planning and spatial cognition.
A few truisms I picked up over my time obsessing about neighborhoods:
- Neighborhood boundaries cannot be right, but they can absolutely be wrong
- Neighborhoods are rarely administratively or politically-defined. As such they bear little/no semblance to traditional admin boundaries (eg Census geography, ZIP, etc…)
- Displaying neighborhood boundaries invites discussion (debate, distraction?) over the boundaries themselves, rather than the intent of finding housing, a restaurant, etc…
Urban Mapping’s approach to neighborhood boundaries has always been rooted in uses that support the needs of local, social and mobile applications from local search to real estate. My team created boundaries using thousands of sources–convention and visitor bureaus, chambers of commerce, the hospitality industry, real estate, city planning, local media, historical conventions and expert knowledge. The broad collection of sources is critical as facts alone cannot answer these questions.
For example, if the community association of the Deep Ellum neighborhood in Dallas doesn’t define boundaries, how can other parties expect to have something more “accurate?” Same goes for local knowledge–if neighborhood boundary data is intended for only residents of (say) Deep Ellum, it makes sense to consider only that constituency. If the intent is to serve a broad base of users–maybe families who relocate, French tourists who visit or martians who emerge from a spaceship– who knows how they will come to understand the meaning of Deep Ellum.
Same applies for Alphabet City, a sub-segment of the East Village in Manhattan. Should it be removed from a definition of NYC neighborhoods because it is not au courant? We say non! It’s all about the user and how data is represented. One reason we think our approach is best is because we consider these differing opinions and recognize the difference between data and applications. In areas of high urban density, some neighborhoods will intersect.
A prime example is SoHo and NoHo in Downtown NYC– Houston St is effectively the dividing line, so if one is on Houston St, does this mean you are in neither neighborhood, SoHo only, NoHo only or both neighborhoods? We choose to represent an overlap of SoHo and NoHo, reflecting the inherent fuzziness. If a point intersects two (or more) neighborhoods, there’s no reason it can’t be associated with multiple neighborhoods.
The GIS-types might start to have convulsions, but don’t mind them. There are always neighborhoods that are culturally dominant–SoHo has a long history and NoHo was ‘born’ in 2003. Data structures can recognize the dominant neighborhood. In this way, the application can tell you the most important neighborhood, but the underlying data doesn’t impair you from thinking otherwise.
Neighborhoods are also functions of density–there are many areas where there simply may not be an informal reference to a given region. With low urban density, there might not be any real reason to define a ‘downtown’ if people don’t refer to the area. Just as large swaths of the world are uninhabited, there are many desolate areas that are not represented by neighborhoods. Larger regions (eg Cascadia, The Midwest, The Thumb) of informal space are probably less useful for navigation, but make for a useful gazetteer.