Java Reference
In-Depth Information
response, this produces $790,000 in revenue. 7 The cost of the
campaign is $600,000 (400,000
$1.50), so the projected profit is
$175,000 ($790K
$15K).
But what if we are more selective? Can we further increase profit?
If we restrict the campaign to the top 30 percent of customers likely to
respond (300,000), we will obtain 70 percent of the likely responders,
or 13,800 responses (20,000
$600K
70
200). Projected revenue is $690,000
(13,800
$50), but cost of the campaign is only $450,000 (300,000
$1.50). So, the projected profit is $225,000 ($690K
$15K). In
this case, we actually increased profits by running a campaign for
fewer customers.
$450K
1.4
Summary
This first chapter discussed how data mining is particularly relevant
to businesses today in solving complex problems. Competition and
the need to improve customer experiences and interactions are
among the motivations, along with a better evaluation of the risks
associated with business processes. We discussed other terms that
are often used for data mining or are related to it. We then introduced
data mining, contrasting it with other forms of advanced analytics
such as OLAP and highlighting a basic process for extracting knowl-
edge and patterns from data. We introduced the notion of a model as a
compact representation of the knowledge or patterns found in the
data. To set the stage for our subsequent discussions, we introduced
typical data mining jargon, which will be revisited in more detail in
later chapters.
As gold mining has been performed through the centuries, so has
it been codified into a repeatable process. Similarly, data mining has
evolved to the stage where the process of mining data has also been
codified. Since data mining has parallels with gold mining, we com-
pared and contrasted the process of data mining with that of gold
mining.
We finished with a discussion on the value of data mining, exploring
reliability as well as a specific example in monetary terms.
7
In an actual modeling, the $50.00 expected profit would be multipled by the
probability of response assigned to each customer. This gives a more precise
expected outcome.
Search WWH ::




Custom Search