Database Reference
In-Depth Information
copied to the file store, and no special transformation is needed) until you
know that you've got something that's ready for analysis? Is the data arriving
too fast at your organization's doorstep for the current analytics platform
to handle? If your answers to any of these questions are “yes,” you need to
consider a Big Data solution.
In Chapter 1, we outlined the characteristics that help you spot Big Data
(velocity, volume, veracity, and variety). We also noted that most of the new
Big Data solutions that you might hear about are more likely to complement,
rather than replace, your current analytics platform (such as a trusted data
warehouse). Together, such complementary approaches can help boost your
company's Big Data IQ.
Let's begin with a good example of how we helped solve a business problem
with Big Data technology. We helped a major company in the financial service
sector (FSS) understand why they were experiencing increasing attrition and
defection rates. We started by examining an entire month's worth of email
(about 40 million messages) from their Customer Service center. In this case,
the volume of stored data wasn't large by any means, but we still used IBM
InfoSphere BigInsights (BigInsights)—IBM's nonforked Hadoop distribu-
tion—because the compute capacity required to perform the analytics was
large. A different approach was needed to chew through the emails quickly
and accurately. Moreover, we didn't know what to expect, so we wanted
maximum flexibility in the scale, types of information, and methods we
could use to performance the analysis. We also didn't know what we would
find, and we wanted the flexibility of acquiring insight by “snapping in”
other data sources; in this case, web browsing behavior, structured accounting
data, and account performance history. It's this sort of flexibility that's often
key in solving problems and moving projects forward. We wanted to start
with this example because it highlights some of the recurring use case pat-
terns that we often see. Specifically this use case provides examples of the
following patterns:
Need to move from a small history (months) to several years
Inclusion of mixed information types, in this case structured account
history along with email
Cross system workflows that required specific file export formatting
that was accomplished in a scripting language that was different than
how the data was originally prepared
Search WWH ::




Custom Search