BACKOFFICE
Strategies and technologies to boost
behind-the scenes efficiencies
“It’s an industry that is just so rich in
information and data that has incredible
predictive powers that we realized we
could put that data to work and harness
it as one among several tools that we
use in the underwriting process,” says
Kevin Toth, SVP and chief underwriting
officer for Harleysville, Pa.-based Harleysville Insurance.
Gaetan Veilleux, senior director of
predictive analytics at Denver-based Valen Technologies Inc., says perhaps 90%
of personal lines carriers now use predictive analytics. “The better companies
are moving in that direction,” Veilleux
says, noting that in commercial lines,
the big companies are using predictive
modeling, the mid-sized carriers are in
flux and the smaller ones are trailing.
The automation underlying predictive analytics also ensures consistency,
says Toth. “We’re able to ensure that accounts and risks with similar characteristics are treated similarly across the
underwriting cycle,” he says. Consistency “was a big part of the reason Harleysville started to go down this road.”
Predictive analytics can automate
much of the underwriting for smaller,
standardized commercial lines policies,
says Julian Pelenur, CTO for Bedford,
Mass.-based vendor FirstBest Systems
Inc. Larger, more complex and more
unusual cases still require the human
touch of an underwriter, he says.
But automation does not make human underwriters obsolete, even when
Akhil Tripathi
Harleysville Insurance
HOW IT WORKS
Predictive analytics works much like the
system described in the book “
Moneyball, The Art of Winning an Unfair
Game,” Diers says, referring to the 2003
bestseller by Michael Lewis.
In “Moneyball,” Lewis describes how
baseball’s Oakland Athletics assembled
winning teams despite a relatively paltry
payroll by de-emphasizing well-known
statistics, such as batting averages and
number of stolen bases. Instead, the A’s
crunched latter-day data with greater predictive power, such as on-base percentage
and slugging percentage, to combine
players who could compensate for each
others’ weaknesses and capitalize on
quirks of opposing pitchers.
Similarly, underwriters who once
awarded a merchant a generic credit for
being in business three years or more, can
now use predictive models to determine
when to begin and end the credit, while
calculating the proper size of the credit at
any point along the timeline, Diers says.
Traditional underwriting views data in
isolation, while predictive analytics examines how risk characteristics operate in
combination with one another, says Toth.
Models can encompass “hundreds and
hundreds” of factors, says Diers.
it takes over some underwriting functions in routine cases, says James McCully, product marketing manager for
San Mateo, Calif.-based Guidewire Software Inc., a vendor of core IT systems.
In fact, observers agree predictive
analytics make underwriters more
valuable to their companies. “It casts
the underwriter’s role as that of a portfolio manager,” working closely with
agencies to evolve products, improve
marketing efficiency and help retain
customers, says Trasancos. Until now,
he says, the role of the underwriter has
been mainly focused on accepting or
rejecting risks.
Diers adds that models should serve
as tools for underwriters, and underwriters should not become subservient
to the models. “There’s definitely a
point where you have to say, ‘I know
this is what the science says, and it just
doesn’t make sense,’” she says.
When human judgment trumps au-
tomation, underwriters essentially are
saying, “I believe this is the right deci-
sion to make for the relationship be-
tween the insured and the insurance
company,” says Diers. “That’s the piece
that you never can automate.”
But carriers can do much to allevi-
ate the need for frequent human inter-
vention on accounts that generate
smaller premiums. To improve the im-
plementation of predictive analytics,
the statisticians and modelers who cre-
ate the tools need a real-world knowl-
edge of commercial lines underwrit-
ing, says Jonathan Oltman, Harleysville
VP of small commercial and predictive
modeling.
“There are carriers out there
that I would call data rich, but
information poor. They’ve got
a large amount of data that’s
available, but they just haven’t
figured out how to harness it in
a usable and serviceable way.”
—Kevin Toth, Harleysville Insurance
