Databases Reference
In-Depth Information
An overall power law
(i.e., a few keywords are
very popular while many
keywords are not
popular at all)
Head
Long Tail il
Keywords
The power law is fractal
(i.e., the power law
repeats itself between
any two points in the
distribution).
Head
The fractal property is
what accounts for
profitable niche
markets within the
overall market vertical.
Long Tail il
Keywords
Figure 3.9. Scalar property of power law distribution.
Foundational Takeaways
•
The selection of query terms by searchers is a component of human information
behavior and is related to human information processing. Therefore, you must view
your selection of keywords from a human information-processing perspective.
At the individual level, we have some guiding heuristics for potential customers
•
as they engage in an information-searching process (e.g., principle of least effort,
principle of uncertainty, and principle of information access).
Using a combination of searcher characteristics, namely the terms they use, we
•
can determine to some degree the underlying intent of searchers. The terms in the
query represent one or more concepts for which the person is searching. As such,
the submission of a query is the start of a communication process.
At the aggregate level, most searcher actions are defined by power law distribu-
•
tions in that there is a small number of items that occur most frequently or have
the biggest impact. Meanwhile a large number of items occur infrequently and
individually have a small impact. This is commonly known as the head and the
tail of the distribution (see Figure 3.9). The combined percentages of the tail can
combine into a significant amount.
Relating Theory to Practice
Selection of keyphrases is the core of the sponsored-search effort. To successfully
select high-performing keyphrases, we must simultaneously focus on the individual
and the entire market segment.
