GW TL
Follow Us
ITA MEMBERSHIP

DATA

DATA

Electronic Chat with Dr. Dan Shoham

Bill Sinn | May 02, 2019

Name: Dr. Dan Shoham

Title: Chief Science Officer

Company: Betterview

Number of years you’ve spent in the insurance business?
3+ years in insurance, 30+ in predictive analytics.

Provide a little background on yourself.
I was born in Israel and moved to the US at age 12. At age 14, I was such a bad student that I was almost kicked out of high school. Fortunately, a Temple University professor saw potential in me and gave me an opportunity to work in his lab over the summer. I felt like the University environment suited me better and soon I started taking courses and dropped out of high school to enter a full-time undergraduate program. Later, I gained a full fellowship and graduated with top honors. I then went to graduate school and earned a Ph.D. in Mathematics at age 21.

My next adventure was as a Staff Scientist at M.I.T. Lincoln Labs in the late 1980s and early 1990s. During my time there, I worked on a variety of far-ahead-of-its-time machine vision technology projects, including the identification of mobile ballistic missile launchers using air- and space-borne imaging radars in the waning days of the Cold War. 

After my work at M.I.T Lincoln Labs, I took an opportunity in the private sector at HNC Software (which was acquired by FICO) developing fraud-reduction algorithms. At HNC Software, I started off as a Senior Scientist, before being promoted to Principal Scientist, and then to Product Line Director. During my tenure there, I was responsible for building predictive analytics products for the financial and payment industry, including the international version of Falcon, a Neural-Network based credit card fraud detection system. Not many people understood back then, in the mid-1990s, the power and potential of Neural Networks. In a matter of a few years, we opened markets in over 60 countries on four continents with a provable fraud-reduction value proposition.

As I was broadening my scope from just the scientific aspects to the full spectrum of connecting the mathematical sciences and the monetization possibilities, I needed to upgrade the business side of my skill set. I returned to school to obtain a joint MBA from Columbia Business School in New York and London Business School, focusing on global management. I also became a licensed California CPA. I obtained these certifications to provide unambiguous credentials to my current and future audiences that I can speak the language of business no less than that of science.

This new skill set was put to use immediately when I founded and became the CEO of Edgeware Analytics, an angel-backed startup providing small business creditworthiness scores. In a few short years, over two billion dollars worth of small business loans were scored and transacted across our online platform. We were able to prove that small business loans that used our scores were, on average, $3,000 more profitable than without our scores. Unfortunately, when the credit markets froze in 2008, lending came to an extended halt and Edgeware Analytics became unprofitable. We still managed to sell the company to a group of investors composed primarily of former clients. The resultant company, Boefly, is still active.

After selling Edgeware Analytics, I had a series of positions, averaging 3 years each, helping establish predictive analytics organizations in operating companies including ARS National, Hyundai Capital America, and Farmers Insurance.

You founded a startup many years ago and then changed to a leading insurance carrier. Now you’re back to an insurtech company. Why the changes?
I feel that the best capacity to contribute comes from having lived on both sides of the vendor-operator divide. To justify original AI work, the value creation has to exceed the risk-adjusted development cost. When I was at Farmers, the value creation was limited to our market share -- about 5% of the US. At Betterview, cutting-edge science has the potential to create value for an entire global market, so the economy of scale enables investments in much more exciting opportunities.

You spent over three years as lead data scientist at Farmers Insurance. What do think was your greatest contribution at that company? 
Without a doubt, my greatest contribution was an auto subrogation model. This model was used to identify subrogation opportunities that claim representatives missed. The model used several new techniques that were never before used at Farmers. Most crucially, a natural language processing algorithm I developed that incorporated text sources (i.e.phone call transcripts, mail correspondence, police reports, etc.) into the overall score. The new model was put into production in a Champion-Challenger mode where a percentage of new claims were scored under the prior system and a percentage under the new score. No one, not myself, not the claim representatives, not even the subrogation people, knew which claims were processed under the old and which under the new system. This was truly a true double-blind environment. Afterward, the total recoveries from both groups were compared. In its first year, the new system generated provable, bottom line, contribution in the tens of millions of dollars compared to total development and deployment costs of well under one million dollars.

How would you describe Betterview? What makes the company unique?
Betterview brings advanced Image Analytics AI to the aerial imagery property evaluation space. In the last few years, after decades of exasperatingly slow progress, image analytics AI has taken off spectacularly, and in some situations, it now outperforms human image comprehension. There is a somewhat rapid dash in many areas to bring these new capabilities to market. Betterview brings the right combination of talents to this unique time point, with a well-rounded set of skills and technologies. My job is to power up the advanced AI component of the story.

What is your vision for your new role at Betterview? What do you hope to change or innovate? 
I would like the Betterview property report to become the baseline standard for the property insurance industry, much like FICO is for consumer credit underwriting. I feel that this could be achieved by making the report very comprehensive, having powerful machine learning image analytics definitively identify important observables and risk factors, and a plethora of risk scores translating these observables into provable dollar-measured value propositions.  Much as credit scoring has become the industry language for lending, I want Betterview visual scoring system to be the equivilant for our markets.

You’ve pioneered a number of analytic models over the years. What changes or challenges do you see with data over the next 5-10 years?
Ohhooo, predicting the future of predictive analytics? I think the biggest change coming is the transition into multi-domain-ness.  It has been an unchallenged predictive analytics conventional wisdom that it is not enough to do good math. In order to build solid models that can also generate leverageable business value, a strong dose of domain knowledge is needed. For example, to build a credit score, you don’t just need a mathematician, you also need a lending domain expert. To build a roof-evaluation score, you also need a roof domain expert. However, as we transition into a future where many different domains are brought together, the challenges multiply. At Betterview we currently bring together insurance, aerial imagery, roofs, software, UI, and AI domain experts.  As more domains are incorporated into the leading models of the coming years, at some point, the complexity of the domains would become overwhelming. No one will be able to be an expert in all the required domains, and teams with a wide spectrum of expertise that do not normally interact with each other will be unmanageable. The challenge will be to develop technologies that can bridge the gap and deliver predictions better than any human while requiring a limited amount of domain expertise.

What do you like most about working with data and analytics? What gets you truly excited about his area of the business?
I love seeing a direct line from predictive power to value creation. There is nothing more joyful than to respond to “it can’t be done” with “yes it can, and I’m going to do it.” With Image analytics AI regularly breaking new “can’t be done” frontiers and Betterview positioned in the right place with the right people, I hope and expect to be saying “yes, it can” very often!

What do you think is the biggest challenge for insurers in the realm of data today?
Where to begin? Perhaps with the risk of becoming outdated. Ironically, insurance, arguably, was the first industry to monetize a form of predictive analytics. Property and casualty insurers were in the business of predicting and statistically pricing losses since at least the 17th century. However, along with the glory of being the first to innovate, comes the misfortune of having the longest time to ossify. Insurance, today, is a very fine and well-established art form that combines law, accounting, and statistics. Transparency and explainability trump predictive performance. Whereas 21st century AI scientists pride themselves on building models so complex that even their own creators don’t really know how they work inside--a necessity for models that outperform humans--actuaries are careful to build models that meet legal, accounting, and regulatory tests. This difference in methods creates a disconnect. It will not be long before data scientists with black box AI systems will develop risk prediction models that outstrip anything legally complying actuaries can produce. Then what? Will the market allow such models, with huge value creation opportunities, to just sit idle on the sidelines? If new entrants challenge the established system by bringing such AI systems into play, how will traditional insurers respond?

What is your passion outside of Betterview?
Outdoors: Mountain hiking. My bragging list includes the Grand Canyon Rim-to-Rim, Mount Whitney in winter conditions, the Inca Trail, Pikes Peak, Half Dome, Telescope Peak, Badwaters-to-Dante’s Peak, and Cactus-to-Clouds at nighttime.

Indoors: Boardgaming. My bragging list includes being the world’s highest-rated Diplomacy player for 5 years (of about 3,000 rated players at the time), and 3 consecutive wins of the California Axis and Allies tournament. 

Do you have any words of advice for today’s insurance IT/business leaders?
Provable dollar-measured value proposition!

 


Featured articles

INSURANCE IT NEWS

Read More

On archive MR

MT MR

ELECTRONIC CHAT

The Email Chat is a regular feature of the ITA Pro magazine and website. We send a series of questions to an insurance IT leader in search of thought-provoking responses on important issues facing the insurance industry.

ITA LIVE 2020

ITA LIVE 2020 –SAVE THE DATE!
April 5th – 7th, 2020
The Diplomat Resort
Hollywood, FL
Become a member today to receive updates – www.itapro.org/MR

BLOGS AND COLUMNS

only online

Only Online Archive

https://www.itapro.org/Register/Membership

ITA Pro Magazine

ITA Pro Magazine

ITA Pro Buyers' Guide

Vendor Views

Partner News

2021 © Insurance Technology Association PRIVACY POLICY TERMS OF SERVICE