PageRank to PlaceRank Is More Than Changing a Few Letters Around

January 5, 2010 · Posted in Commentary, Companies

There was a great article yesterday by Chris Silvery, who works for search engine marketing firm Key Relevance and is a regular contributor to the Local’s Only Section of Search Engine Land. The article highlights some of the ways that location oriented search within Google behaves, and frankly how it very often doesn’t behave the way it ‘should’.

Per John Hanke, VP of Google Earth, Maps, and Local from a recent TechCrunch article : ”PlaceRank is like PageRank for places, it tries to figure out how prominent a place is based on factors such as references on the Web, reviews, photos, how many people know about it, how long its been around.”

By the way I think it’s notable that the thing being “figured out” here is “prominence”.

Now I understand that you’ve got to start somewhere, but I would argue that the tactics used in web search engines don’t really apply to spatial search much and it should be treated as a completely separate animal. Web pages are about text and the authors of that text linking to (and as a result voting on) other web pages, in order to determine a pages’ ‘prominence’.

I think there are a few key differences when looking at location and spatial oriented search:

The “prominence” of a search result is relative to things like distance and the convenience of alternatives in local/spatial search, versus something more absolute in web search where you’re simply clicking on a link to ‘get there’
Determining “prominence” is very important when parsing through 1 trillion pages of “always available” information, but in the more dynamic yet much more limited options of local search something as simple as solving for “highest prominence” may not be the right answer
The true “linking” happening to a physical place is not happening on a website, but through foot traffic and phone calls… and the traffic links between places is not captured on a webpage at all, but on a handset or a carriers’ back end logs
The stuff being searched for could and should exist in a variety of mediums, not just html on webservers… find a person from their mobile device, find an item from an inventory system, find a bus from a location sensor.

I suspect that there will be some incremental improvements over time with matching online information to offline stuff, but I also suspect that we’d be better off by blowing up the existing search model and starting over from the ground up with a model designed purely around location specific spatial search, that merely taps into the vast reservoir of online content only when necessary… rather than serving as the foundation.

Bookmark It