At MassMutual’s data science lab, members of a developmental program work side by side with company veterans on complex analytics projects. As part of our recent visit to the lab, INN sat down with a typical MassMutual team to discuss a particular project, how the business side works with the lab, and the opportunities for young data professionals in insurance. Participating in this roundtable discussion were:
- Sherriff Balogun, Business Consultant: Sherriff is the main liason between the business and the lab, splitting his time between the lab in Amherst and the main MassMutual campus in Springfield.
- Marc Maier, Data Scientist and Development Program Head: Marc works at the lab full-time on data projects and with developmental program members.
- Hayley Carlotto, Development Program Member: Currently pursuing a master’s degree in computer science.
- Yue Tang, Development Program Member: Currently pursuing a master’s degree in statistics.
INN: Tell us a little bit about the project.
Sherriff Balogun: We are collaborating with corporate compliance and the anti-money laundering group. They track different kinds of transactions and have rules in place to evaluate the risk of money laundering, but want to add rigor to that process. They have four to five data sets they look at, but it’s not a very rigorous process for identifying the risk of any particular transaction. Our job is to consolidate a lot of those data sources together, working on a model where we can look forward into the future with reasonable probability.
INN: What are some of the different data sources that you use?
Marc Maier: We get access to large batches of static historical data simultaneous to building the processes to consume new data. With historical data we can do back testing, and then the pilots can go live with transactions that are going on.
Balogun: Having that historical data gives us something to begin with, and we also get access to new data sources by engaging different parts of the business.
INN: What is the typical cycle for this kind of project?
Balogun: The first phase we collecting data. Then phase two is the meat, where Marc and the data science team will do all the modeling. Phase three is testing.
Maier: As data science lead I’ll work two to four projects at a time; developmental program members are on one or maybe one and a half depending on the complexity of the project and balancing their education. We are probably a couple weeks away from deploying and piloting this model.
INN: What are some of the challenges the business has faced?
Maier: The number of policies and transactions that are coming in is growing because MM is growing as a company. We have to make sure this set of rules is comprehensive enough. In order to really reconstruct the compliance process, we need to figure out what features we want to use. There’s been a lot of data source wrangling.
INN: What is the end goal of this project for solving that?
Balogun: This is mainly for the end business user. That scoring from their perspective is a prioritization of their work. Engagement with the business is giving them vision into the black box of data science.
INN: For the developmental team members: How have you liked your work so far?
Yue Tang: For the entire month of June we had workshops, with team members teaching us skills to make sure we’re in a spot to start on new projects. July we did a warm-up project.
Hayley Carlotto: This is our first major project. We just said the other day that this project allowed us to touch on some data science skills. Each step we were learning something new.
