In addition to simulations, it is often important to have a reporting
and/or analysis capability with your personalization scheme to
alert the developer at how successful certain implementations
or rules are, and whether or not to change them. Simulations
run beforehand to predict, and reports run during or
after to tell you whether your prediction was a good one.
In addition to knowing what rules you want to push, you need
to decide what information you need from the user to supplement
your schemes. You can collect many many megabytes of information
from a single user visit, so it is important to make sure
you don't overload by taking too much data. Decide what you think
you need to collect, and then cut half of them - it's said that the
simplest solution is always the best, and for collecting data,
the simpler the information you need to collect, the better.
Next time, we'll look at some actual code and implementations
of personalization schemes.
