A good analytic model implementation or good analytic strategy is like your
favorite meal. Thinking through such a pleasurable experience might help put
you in an open frame of mind.
Close your eyes and think of one of your favorite meals. Take in all your
senses and imagine why it's so high on your list. What does it look like?
What's in it? How does it make you feel before you eat it, during the meal,
and after the meal? What's the main ingredient? What other ingredients make
it stand out? Do you like the ingredients? Who would you share it with (their
own portion, of course)? To whom would you recommend it? Is this a special
occasion meal or would you make it a regular staple?
The Appetite for Analytics
Think about what goes into making a good meal.
- The ingredients need to be familiar to the diner. Tofu just doesn't
work for some people. Crickets (note Andrew Zimmer)? There are degrees of
familiarity within our habitual walls.
- The ingredients need to blend together and complement each other. Balance
of flavors is critical. Some of these combinations may seem unusual to the
average person.
- The meal should be served the way the diner is used to having it
served—corn on the cob or cut off the cob and served as kernels?
- Should there be changes to traditional dishes (Rueben sandwich with
hummus or Hatch chilies)?
- Are the ingredients and preparation considering a healthy lifestyle? Are
they vegan or gluten free?
- Is the food preparer a food lover who is passionate enough to please
those who will be eating what is prepared? Is the preparer open to
substitution based on allergies, geography, phobias, religious beliefs,
etc.?
Chef de Cuisine
Meal preparation professionals know these requirements from training and
experience. Most of them probably burned a loaf of bread, made a soggy batch of
rice, or undercooked a nice cut of meat at some point. But they also go through
several career steps, gaining insights from mentors and coworkers along the
way.
Similarly, analytic leaders, including C-Suite management, who are
championing an analytics culture, have to be aware but often don't get an
analytics mentor/student experience. Perhaps some in management came from a
quantitative background. But more than likely, leaders probably haven't had
the complete experience of conceiving the meal, creating the recipe, assembling
the ingredients, or doing the meal preparation. Without assistance or
experience, the diner might just shuffle the food around the plate before
ultimately pushing it aside. There may be poor Yelp or Urbanspoon reviews. The
worst would be an unpopular and then closed restaurant or unanswered dinner
invitations by friends and relatives.
Recipe for Success
The planning and recipe is a blueprint for success of creating anticipation
of needs being met, a pleasing experience during the meal, and the resultant
primal comfort in having had the experience. It's even better if the diners
express their wish for an opportunity for a return visit. The mere recall
should make their mouths water.
Like the perfect meal experience, the analytic experience can be thought of
in the same light for your organization. Everything is prepared for and used to
fit the ultimate business purpose—to make better decisions, make them faster,
and make them at less cost.
- Was the strategy well communicated from the beginning? (Was it
taste-tested in advance?)
- Is the process driven by relevant data that is consistently available,
and does it make intuitive sense for the purpose (just like the ingredients
of a recipe)?
- Is the strategy and tool "owned" by the users of the process?
(Did they get to decide what's served and with which side dishes?)
- Is it understood by the affected downstream entities who are
incorporating the processes? (Is there transparency as to how the meal was
prepared (no secret crickets)?)
- Do all stakeholders have clear and well-communicated returns on
investments to users along the process chain? (Are the ingredients healthy
and nutritious?)
- Are the purpose and scope clearly defined by management? (The chef loves
cooking as much as eating.)
- Are the analytics complementary to the overall business strategy? (Does
white wine pair with prime rib?)
Humalogy™
Humalogy the word used to describe the blending of humans and technology.
(See http://www.fpov.com/humalogy-overview/.) Most processes and
actions involve some human effort that is aided by technology. The trick is
figuring out how much human effort versus technology effort needs to be used to
maximize the efficiency of the action.
This is becoming a real challenge for insurers as they strive to interject
analytics into the decision process. One school of thought is to bring new
levels of skill to positions, such as underwriters, claim adjustors, or even
in-house agents. Insurers are gearing up to train their current staff to be
data scientists, whereby the person is expected to understand how to sort
through complex arrays of data and indications as they make decisions
they're used to making based on experience and traditional risk
characteristics, particularly in the medium to large commercial lines
accounts.
New studies indicate underwriters will be expected to rely more on
data-driven insights and less on traditional underwriting theory, consuming
data through analytics to be more efficient and providing bandwidth to enhance
their roles. What underwriter wouldn't want to fill the role of sales
executive, decision scientist, customer advocate, and innovator? Do such skill
sets exist in our underwriting teams? If not, what is the cost and ramp-up
investment to get them there? Do they want to fill this new style of
underwriter? Will the next generation of underwriter arrive at insurers'
doorsteps with all of these skills in hand? Does the recipe for more edible
analytics support this trend?
Chief Cook and Bottle Washer?
These new underwriting skills or needs sound much like those of a data
scientist. In 2016, some studies indicated the job position of data scientist
was the best job to have. The pay is lucrative, and good data scientists could
write their own ticket to nearly any company where they might want to work.
Being the most coveted, the role has a pay scale to match. The requirements of
a data scientist include being adept at understanding data, versed in the
nuances of analytical techniques used, and a master at data visualization to
communicate what the data is telling them. All this needs to be paired with
deep insurance domain expertise for such skills to fit the demands of
underwriter along with the data scientist role.
How do we strike a balance? Can a work colleague be successful or be
proficient in all the skills of the maître d', chef, waitperson, sous chef,
wine steward, dessert waiter, dish washer, and cashier? One could argue this
type of person is more of a unicorn—able to juggle both technical and human
skills.
Citizen Data Scientist
Maybe the future of underwriting and the use of analytic tools won't be
such a revolution with skill requirements being as technical as some think.
With new machine learning and your computer, most of us can be self-taught data
scientists. Almost anyone can find the complex interactions of variables and
the relevant predictive variables (and combinations) using Natural Language
Processing (NLP) that is now being rolled out.
NLP connects human communication to the computer using voice synthesizing,
voice recognition, and past interactions to understand what the user is seeking
to understand. Therefore, unlike a data scientist having to be versed in using
analytic tools such as SQL, R, or SAS, the problem can be communicated and
understood by the computer and then set on its way to provide various inputs
based on what the data says. Skeptics of the potential reality of these
technologies should ask Wilson (of IBM).
Food for Thought
My mouth waters every time I sit down to write or edit this paper. It waters
not only because I love good food, but also because you can call me late, but
you cannot call me late for dinner. I just follow my nose and my gut when
challenged with providing advice on implementing analytics into different parts
of the organization.
There's a lot to think about and several decisions to make as a company
to develop a data-driven culture. Thinking through the entire meal preparation
process can help. You can rest assured, if you don't deeply consider all
facets of the journey, it will eat your lunch.