Location is fundamental to our personal and business lives – yet it is a concept that is easily and often misunderstood. The Oxford Dictionary defines this word as “the point or extent of space that is occupied by a person, place or thing”. So a place is not a location – it has a location…
In fact, location is an abstract concept that is described by geometry (e.g. lat/long coordinates, boundary polygon, or the intersection of two lines for a major street intersection). This geometry, when used with a geographic reference system, can define “where” in the world a person, place or thing is located. And in turn, mapping systems can depict this location at various levels of granularity or zoom scale.
Most human beings have spatial intelligence and some degree of geographic knowledge – so the location encoded in the answer “near San Jose” to the question “where are you?” actually means something. The vast majority of computers systems (with the exception of mapping and GIS systems) might achieve a string match on San Jose, but will have no concept of near – or closest, adjacent, contained by, or within 5 miles.
Invariably, the answer to the question where is a place reference (except in the military, where “geographic coordinates” are commonly used). This reference could be a street address, building name, intersection, neighborhood, district, region, natural or man made landmark, city, state, country, and numerous other concepts for place. These are all physical real world objects, unlike location, which describes a theoretical spatial extent.
Places are identified by addresses, place names (often building occupant or owners name) and other codes – and always have a location. Within enterprise computing and the web, location context is emerging as an important way to organize and find information -but location it is really a function of place. In a search-oriented user interface, a simple “where are you looking for it?” box could prompt a wide range of place references at many levels of granularity, from country to a suite within a building or store within a mall.
According to an urban myth (some city or county level study, completed 15 years ago that is referenced by virtually everyone involved with maps and GIS), 80% or more of the worlds information is associated with locations. Actually, it’s not! This content contains place references such as addresses or place names, and must be transformed by a process called geocoding (determine the lat/long coordinates for street address, ZIP code or city name), to associate the content with a location.
Why does this distinction matter? Simply – if this content is not enhanced with geometry and geographic data elements, and linked to and indexed by a spatially-enabled search or mapping engine, there is no location intelligence! In an earlier part of this series, I pointed out that there is a tiny amount of content indexed by Local Search engines, as compared to the keyword search engines in the Web. This place-to-location conundrum, at the data level, is reason.
So, we have established that places are real world objects and always have a location. What about the other objects we represent in our databases and documents… people and things? This gets interesting because places typically don’t move around, but people and things do! In fact, according to United States Postal Service and directory publishers like Yellow Pages, 25 – 30 % of consumers and businesses move every year. But, using the model we have established, they are not moving (directly) to new locations, but rather to a new physical home, apartment or place of business – which has a location. Errors, discrepancies, and lack of standards in address entry interfaces make address data integrity and thus, the accuracy of location a global problem.
And finally, we have vehicles, cell phones, or PDAs with GPS or other “locator” technologies… these mobile things (or the person carrying them) are not directly associated with a physical place. Mobile assets and resources have a “position” and may be “at” or near place.
Mapping engines, spatial indexes, and GeoTags (geocode tags on content) have center stage in Local Search at the moment, but there are several big problems with this approach:
1. Precision – there are 20 to 50 meter error factors involved in the various address validation, geocoding, map data, spatial index, GPS and mapping components needed to determine and display a location. A lat/long pair does not uniquely identify a building, let alone a suite or store within a building.
2. The vast majority of business data and web content contains place references and not location attributes, and must be processed, indexed, integrated and linked to a spatially-enabled search/mapping engine to enable location-centric search.
3. There is a hierarchy of places (e.g. suite, building, street, city, county, state, country), and the relevance of the search result can vary widely based on where in this heirarchy, the user initially anchors his search. In other words, a single piece of content could be (appropriately) associated with all the places in this list.
4. Reverse Geocode – determining the real world place from a lat/long coordinate, say from a GPS chip on a cell phone, is fraught with problems and is highly inaccurate, due to the precision issues above and the lack of a definitive database of places, with precise lat/long attributes, to compare the position to.
A use case that ties this all together, is inventory on the shelf of a specific store. The Mall of America in Bloomington MN, has 9 different valid street addresses (and associated lat/long attributes) – none of which uniquely identify any one of 520 stores. Items in the store are associated with specific places, but couldn’t be geocoded, mapped and located in a Local Search context, until they are linked to the higher level mall object, and associated with the “nearest” mall entrance, based on a detailed floor plan.
A URL for a virtual place in the Internet is unique, precise and reliable for finding content, regardless how far down the tree structure of specific domain name, the content is organized. This concept of a unique place identifier, place hierarchy and index does not exist in the way content about real world places is currently organized.
If indeed 80% of the worlds content is associated with places, and these places all have locations, then a universal method to index and organize this content, based on place-centricity, and associated location intelligence, would be major step forward for the Local Search and Location-Bases Services (LBS) industry.