What's Your StORy?

John Toczek

John Toczek

June 2016 What's Your StORy?
AVP Predictive Modeling & Analytics at Chubb

More questions for John Toczek?
Check out the Open Forum on INFORMS Connect!


What has been your best INFORMS experience so far?
I love the Business Analytics conferences and have been to all of them in the past 10 years. The talks are always great but the camaraderie is what I look forward to most. Some people you only see once a year but I’ve made some great friendships through those interactions. There is a ton of excitement for our field and I’ve been fortunate to have a front row seat to witness the growth.


What is your favorite O.R. application?
I love using simulation to solve problems. Elegant solutions have their place but every once in a while it’s fun to break out the sledgehammer.


How did you become “the PuzzlOR”?
I blindly sent an email to Peter Horner (Editor, OR/MS Today & Analytics) in January 2008 that said, “I've been kicking around an idea for a new section in OR/MS Today called the PuzzlOR. It would be a relatively easy problem that requires O.R. techniques to solve.” Peter replied, “We've never published anything like this, but I think it's crazy enough to work.” It struck a chord with the greater O.R. community and I think that is because when you distill what we do down to its essence, we all love puzzles. It’s hard to believe I’ve been writing these for eight years now.


What recent project have you been involved in that you are proud of?
There are so many exciting projects going on right now in the Global Analytics Predictive Modeling group at Chubb insurance. I’m in the middle of a project to predict automobile accident fraud in Mexico. And I’m working on some rate modeling for a life insurance project in Thailand. Did you know you can buy life insurance at car dealerships there? The cultural dimensions add another level of fascination when building and implementing models.


What is something you learned in the last month?
I’m learning about unsupervised text mining, which allows you to identify words and phrases as binary variables that are correlated with a target variable. This is a new area for me that I haven’t had much exposure to. It’s an interesting topic that is becoming more and more popular. Now that companies are getting better at using their structured data, the next horizon is the unstructured data like text. Also, I learned which of the three elevators in my building will reach the first floor first by modeling them over time. The elevators have an indicator above that shows the floor they are on and the direction they are traveling. This is the topic of the next PuzzlOR, so stay tuned!


If you were a Microsoft Office program, which one would you be?
I’m Clippy but I’ve been upgraded with neural network learning algorithms so I’m actually good at helping you now.


What about your career might surprise us?
Like many in the O.R. profession, I took the long road to get here. I started out as a chemical engineer working in steel plants.


Where’s Waldo?
I was never very good at these but Randal Olson did a fun analysis on this question back in February 2015. He used kernel density estimates to come up with an optimal search strategy that will find Waldo in the least amount of time. Essentially you’ll want to start in the top left of the page and follow the path of an “S” that has been tilted to the left. It’s a great short read you can find online.


Here’s a puzzle for you: There are a dozen eggs in a carton. Twelve people each take a single egg, but there is one egg left in the carton. How?
There was a tiny chicken in the carton and she laid a 13th egg while the other 12 were being taken. (Either that or it was a baker’s dozen.)