Features:
INFORMATION INTO ACTION
by J. William Bell, NCSA
Allstate Insurance Company wins the NCSA Private Sector Program's 2003 Grand
Challenge Award, honoring more than five years of data-mining innovation.
There's value to be found in any pile of data. When you're facing a molehill,
it's easier to dig up. Patterns, if they exist, can be found with little
effort. Predictions can be made based on those patterns. And any deviation
from the norm catches your eye. But for Allstate Insurance Company, molehills
gave way to mountains not long after the company paid its first claim on a
broken car door handle in 1931. Today, with more than 30 million policies in
force, finding patterns and deviations in the vast customer, geographic,
economic, and other data requires more than just a keen actuary and a slide
rule.
To glean more insight from the data that they collect and have access to,
Allstate has worked with members of the NCSA's Automated Learning Group for
more than five years in a field known as data mining. Data mining combines
statistics, machine learning, pattern recognition, database management, and
high-performance computing to produce automated methods for analyzing data
sets and discovering new knowledge. It's a burgeoning and challenging field
that no company can afford to ignore in today's rapidly evolving business
environment.
Together, the team has taken advantage of data mining to improve customer
service, experiment with new pricing systems, and develop better ways for the
company's analysts to keep up with changes in the competitive environment.
These efforts allow the company to work more efficiently and better target
their products to their customers -- whether those customers are buying a new
car, insuring their first home, or recovering from a fire.
In April NCSA's Private Sector Program presented its 2003 Grand Challenge
Award to Allstate, honoring this expansive data-mining relationship.
Wide range of implications
NCSA's Private Sector Program gives partners access to all of NCSA's leading-
edge technology and knowledge. It creates an environment in which companies
are free to experiment with their most daring ideas. The annual Grand
Challenge Award honors breakthrough research completed by private sector
partners while working with NCSA. These breakthroughs ensure America's
leadership in global business.
"Our relationship with NCSA is…judged on one critical dimension and that is
our ability to execute our objectives in the marketplace better than anyone
else. [We're going to] be able to take reams and reams of customer data, and
reams and reams of customer transactions, and turn those into actions that the
customers value," says Ronald McNeil, Allstate's senior vice president of
product operations. "The partnership between NCSA and Allstate has allowed us
to turn information into action quicker than our competitors."
NCSA's Director Dan Reed says that turning information into action has wide-
ranging implications. "Many of the problems in data management and data mining
cut across a broad range of business environments and business sectors. The
common theme is business intelligence. And the way you glean that intelligence
is by taking advantage of the raw data that your business produces in everyday
practice. Data mining…gives you the power to be earlier to market, to shape
products that more correctly and more accurately match the expectations of
your customer base. It allows you to tailor products, to respond more nimbly
to the business environment."
More than 100 data-mining modules
Allstate's work with NCSA relies heavily on the Automated Learning Group's
premiere software, called D2K. Short for Data to Knowledge, D2K integrates
more than 100 modules representing both common and unique data-mining
algorithms. These algorithms can, among other things, clean up data sets and
prepare them for computations, search for patterns, make predictions, identify
unusual features, and visualize the data for further analysis. Within a visual
programming environment, users can connect these modules and create powerful
problem-solving systems.
From the data, they can forge knowledge. Within a visual programming
environment, users can connect these modules and create powerful problem-
solving systems. From the data, they can forge knowledge. And from that
knowledge companies like Allstate can make better decisions--about who and
what to insure, how much to charge, and how to deliver what their customers
may want and need.
"NCSA is one of the only places in the world that can offer a computational
resource--the TeraGrid--a data-mining framework--D2K--and access to
application expertise through the Automated Learning Group to address these
large-scale data-mining applications," says Michael Welge, director of the
Automated Learning Group.
Tactical Competitive Intelligence Network
One of Allstate's greatest data-mining successes has come with their Tactical
Competitive Intelligence Network, or TCIN. Relying on a subset of D2K modules
that represent sophisticated text-analysis algorithms, TCIN is a boon for the
company's analysts. Analysts search the Internet and other information sources
for data on things like competitors' pricing structures or the safety ratings
of a particular make of car, then they refine the cost of insurance policies
offered by the company.
The TCIN system relies on another piece of NCSA software, called VIAS, to
retrieve data from the Web, newsgroups, and public mailing lists to analyze
competitive information. Algorithms selected from the suite of D2K tools and
easily combined using D2K's visual programming environment then filter the
significant portions of the data retrieved, group similar and related pieces
of information, and classify those groupings. The system even visualizes the
results. And all of this is done automatically.
According to Jeff Deigl, Allstate's assistant vice president of product
operations, research and development, analysts use "three quarters of the time
spent on a project getting the data and getting it prepared." Only that last
25 percent is spent understanding what the data's telling them. "By shrinking
the time that the analyst needs to spend on data prep, it allows them to get
to more projects and to be more effective in the analysis end product. Which
is where, really, the key decisions are made."
Territorial Rate Making
Allstate's Territorial Rate Making project gives the company another
opportunity to refine the methods used to set the price of insurance policies
in a given region. Traditionally, companies have simply looked at the
historical data for a given territory. With the help of NCSA and D2K, however,
Allstate is taking a more complicated view.
"[I]n setting our territorial rates, we've relied only on our internal loss
data to estimate what we should charge in different geographical
locations…NCSA and their tools have allowed us to supplement that information
with external data," says Deigl. "For example, in assessing the fire exposure
we face, understanding the forest cover and the amount of precipitation in an
area is very useful and helps us better estimate the losses we'll incur."
By combining all of this data--and more importantly, by drawing new knowledge
from this data--Allstate is doing what it's always done best: Delivering
competitively priced products that help meet the needs of our customers. It's
a simple goal. And, through Allstate's partnership with NCSA's Private Sector
Program, it's a goal that's better realized every day.
"Our relationship with NCSA has grown over the years, and I would consider it
now to be an extraordinary partnership. NCSA provides us a research
environment that includes leading-edge hardware, software, and analytic
methods that, when coupled with the research staff, allows us to address
Allstate business problems in ways that we can't in our own computing
environment," says Roger Einbecker, Allstate's assistant vice president of
information technology.
"NCSA helps us to better understand the knowledge of our business that's
locked in our data."
Relevant URLs
|
