Page 405 -

Chapter 9 Business Intelligence Systems
404
Metadata

BI Application

Push
BI Data BI BI BI Server “Any”
Source Application Application Device
Result Pull
BI users

• Operational data • RFM • Computer
• Data warehouse • OLAP • Mobile devices
• Data mart • Other reports • Office and other applications
• Content material • Market basket • Cloud services to anything...
• Human interviews • Decision tree
• Other data mining
• Content indexing
• RSS feed
• Expert system
BI System
Figure 9-29
Elements of a BI System
BI servers use metadata to determine what results to send to which users and, possibly, on
which schedule. Today, the expectation is that BI results can be delivered to “any” device. In prac-
tice, any is interpreted to mean computers, smartphones, tablets, applications such as Microsoft
Office, and SOA Web services.

Q9-9 2026?

BI systems truly add value. As described in the Guide on pages 408–409, not every system is a
success, but simple ones like RFM and OLAP often are, and even complicated and expensive data
mining applications can generate tremendous return if they are applied to appropriate problems
and are well designed and implemented.
For example, suppose you never buy expensive jewelry on your credit card. If you travel to
South America and attempt to buy a $5,000 diamond bracelet using that credit card, watch what
happens! Especially if you make the attempt on a credit card other than the one for which you
paid for the travel. A data mining application integrated into the credit card agency’s purchase-
approval process will detect the unusual pattern, on the spot, and require you to personally verify
the purchase on the telephone or in some other way before it will accept the charge. Such applica-
tions are exceedingly accurate because they are well designed and implemented by some of the
world’s best data miners.
How will this change by 2026? We know that data storage is free, that CPU processors are
becoming nearly so, that the world is generating and storing exponentially more information
about customers, and that data mining techniques are only going to get better. It is likely that by
2026 some companies will know more about your purchasing psyche than you, your mother, or
your analyst.
In fact, it may be important to ask the question: How unsupervised do we want unsupervised
data mining to be? Today, a data miner extracts a data set and inputs it into an unsupervised data