BISA Magazine - Quarter 1, 2017 - 24

FEATURE STORY

cal history, no moving violations, and
no crazy hobbies ... You need no more
underwriting. You're done." Accepted.
Or alternatively, "If you would just run
down the block and have the pharmacy
on the corner take your blood pressure ...Their system is linked with our
system." That is, a quick detour and
presto: Approval.
Lucker is not wholly sanguine vis-à-vis
this scenario, however. Predictive analytics relies on a lot of public data-e.g.,
did you get a speeding ticket recently?
Are you married? Did you vote in the
last election? It also relies on demographic information, much of it provided by so-called data brokers, i.e.,
non-public concerns like Acxiom, Experian, Epsilon, and Datalogix. Insurers
that use PA often base their underwriting decisions on these sorts of primary
and secondary data sources, and there's
no easy way to correct that information
when it is wrong.
Just look at residential home data-the
sort your town collects. This information can be rife with errors. In the
past, the fact that your town office says
(falsely) that you have a slate roof atop
your house might not matter one whit,
but if an insurer is also using that database, and ultimately using it to make
decisions on your homeowners insur-

24
www.bisanet.org

ance policy, how do you get the town
to care enough to correct its mistake?
(The insurer, for its part, is not going
to send someone out by car to visually
confirm your roof claim-it's not worth
the insurer's time.)
In addition, data brokers, whose data
may be accurate in the aggregate,
continue to commit mistakes on
the individual level. Lucker recently
asked 87 people on his Deloitte
Consulting team to look at their own
records held by a data broker. Half of
the 87 reported their personal data
contained errors; one participant born
in 1987, for example, was listed as
69 years old. Among the comments
of the participants who accessed
their records:
* "It said I am a high school graduate
(I have a PhD) and that I'm married
(I am not)."
* "I have owned the same home
for over nine years; they show no
record of it."
* "I wish this was all I spent in a
year!" [The data report listed the individual's yearly expenses at $458.]
* "I appear to be interested in absolutely every category-sewing ... really?"

A persistent question regarding this
sort of information is: "How do you
fix it when it's wrong?" says Lucker. In
other words, how do you amend those
public records, consumer marketing
data, warranty information, health data,
and so on?
Companies need to create linkages
back to primary data sources, he continues. There's this notion that the data
is available, accurate, and ready to be
used, but it isn't.
Think of all the junk mail you receive
with your name misspelled. There is still
no self-righting mechanism. There is no
easy, comprehensive way to notify data
brokers or companies that your data is
wrong. Even if you notify one data broker, another one will still have the false
information. In the example of an incorrectly identified house roof, there is no
way for the data broker to tell township
XYZ that the information is wrong.
Lucker notes insurers that are moving
forward with PA are not really getting
more accurate data, rather they are
doing a sort of cost-benefit analysis;
if they skip "fluids" (i.e, taking blood
and urine samples) and use predictive
analytics metrics instead, they will incur
more mortality risk (not less), but it will
still be less costly than sending a nurse
to your house to take fluids and an EKG.

http://www.bisanet.org

Table of Contents for the Digital Edition of BISA Magazine - Quarter 1, 2017