Databases Reference
In-Depth Information
as a surrogate for relevance of a result to a query). Fourth, with sponsored-search
logs, the users may be aware that they are being recorded and may alter their actions.
Therefore, the user behaviors may not be altogether natural.
SSA also suffers from shortcoming deriving from the characteristics of the data
collection. Hilbert and Redmiles [
43
] maintain that all research methods suffer from
some combination of abstraction, selection, reduction, context, and evolution problems
that limit scalability and quality of results. SSA suffers from these five shortcomings:
•
Abstraction problem -
how does low-level data relate to higher-level concepts?
•
Selection problem -
how does one separate the necessary data from the unneces-
sary data prior to reporting and analysis?
•
Reduction problem -
how does one reduce the complexity and size of the data set
prior to reporting and analysis?
•
Context problem -
how does one interpret the significance of events or states
within state chains?
•
Evolution problem -
how can one alter data-collection applications without
impacting application deployment or use?
Potpourri
: Many aspects of Web analytics can be difficult with many caveats and
potential pitfalls. One classic example is known as ”the hotel problem,” named by
and credited to Rufus Evison [
44
].
The hotel problem is used as an example to show the effect that the date range
has on Web analytics results, and that comparing results between different data
ranges can cause seemingly nonsensical measurements.
The hotel analogy is a simple way to illustrate this point, by showing that the
unique visitors for each day in a week might not add up to the same total as the
unique visitors for that week. (Note: It could be day to week, week to month,
month to year, or whatever.)
The hotel problem basically goes like this.
A hotel has two rooms. Each room has a guest each day during the week.
Therefore, the unique visitors per day are two.
One might think that to get the unique visitors for the week, you just add them up
for the seven days, which would be fourteen, assuming the hotel is full each day.
However, this methodology is flawed. Why?
What if one guest stayed in the hotel for seven days? This guest would be
counted as a unique visitor each individual day but only once when counting
unique visitors for the entire week.
So, assuming one guest stayed in a room the entire seven days, and the other
room had a new guest each day, our unique visitors for each day would be two.
Our unique visitors for all seven days would be eight.
Because each method has its own combination of abstraction, selection, reduc-
tion, context, and evolution problems, there is a need for complementary methods of
data collection and analysis.
