More on Categorization and Local Search Results
Last time we mentioned that there were an alarming number of variables in the local search categorization stew. We discussed several, but there are two left before we can get a better idea of why so many local searches have flawed results due to categorization issues.
First, it is clear that users of local search categorize their input by choosing words to describe their target of interest. Obviously many of these choices are responsible for producing results that do not satisfy the search. But for now, let’s assume that the user provides a common, descriptive and unambiguous set of descriptors to describe their target of interest.
Next, local search providers usually pre-categorize their business listings database into categories of business that they feel best reflect the terms their audience uses to find specific business types. They index these categories by keywords, brand names, SIC, NAICS, Yellow Page headings and other cues. Schematically, it looks something like this
In essence, the local search provider decides which ingredients belong in a specific category. Local search providers constantly re-evaluate their “internal categorizations” by mining their search logs and evaluating the quality of the results. Although the logs do not contain a specific indication of the user’s satisfaction with the results, the records do provide surrogate evidence of dissatisfaction, such as the user choosing to ignore the result and recasting the search or abandoning it completely.
Local search providers believe that their work related to categorization of business types is the “secret sauce” of their success. I agree, although many of the “sauces” being used seem to need more flavoring.
See the following three examples of categorization when using local search to find an “electric range”.
Clearly, I was looking for something specific (an “electric range” and each local search provider (Google, Yahoo and YellowPages.com) slotted my search terms into their categorization, which then was used to extract the listings from their database that matched the intent and geography of my query based on the parameters and classifications used by each system. Clearly Google and Yellowpages.com matched up well. Yahoo? Well, those are some interesting results.
Well, we have run out of space again. I promise I’ll finish this off next time.