KDnuggets Home » News » 2014 » Jun » Opinions, Interviews » Interview: Tom Kern, Risk Modeling Manager, Paychex on Risk Analytics and Sales Anticipation Model ( 16:n16 )

Interview: Tom Kern, Risk Modeling Manager, Paychex on Risk Analytics and Sales Anticipation Model


We discuss the role of Risk Analytics at Paychex, strategic importance of Sales Anticipation Model, optimizing business processes by leveraging Big Data, and advice for companies thinking about Big Data as well as aspiring students.



Tom KernTom Kern is a Risk Modeling Manager at Paychex, Inc. Under the Risk Management umbrella, Tom helps to coordinate and execute a wide range of projects centered around predictive modeling, optimizing processes in all departments from sales strategy to internal operations and mitigating risk throughout the company.

Tom joined Paychex in 2012. Prior to Paychex, Tom was a Predictive Modeling Analyst with a large digital marketing agency, servicing major clients in the financial services, insurance, and automotive industries. He holds a MA from Boston University in Applied Statistics, and a BA from Boston University in Applied Mathematics.

Here is my interview with him:

Anmol Rajpurohit: Q1. How would you define "Risk Analytics"? How have your Risk Models changed over last few years?

Tom Kern: For many companies, Paychex included, Risk Analytics has moved beyond a traditional value conservation framework. Risk AnalyticsToday, Risk Analytics often encompasses data-driven intelligence that informs a nearly exhaustive list of enterprise-wide decision making, furthering value creation at least as often as value preservation. Expansion of technological capabilities and analytical expertise, along with an evolving complexity of business needs, have augmented the sophistication of the risk models we have produced in the last several years. We explore a wider array of modeling techniques and look for deeper methods for direct integration of model scores in database systems and operational processes.

AR: Q2. What was the strategic imperative behind building the Sales Anticipation Model? How did you go about building this model and integrating it into firm's decision making?

TK: Incorporating greater intelligence into considerations of the sales organization positioned the Risk Analytics team to exert a much greater level of influence than had previously been enjoyed. Considering the size of our sales organization, helping to better inform the location, distribution, focus, and expectations of sales reps showed clear and immediate opportunity for revenue maximizations. As this was our first concrete collaboration with the sales organization, we encouraged a very high level of end user participation in the construction and deployment of the Sales Anticipation Model. In that way, the model became a shared vision that garnered much greater support than were it to be created in a silo.

AR: Q3. How can Big Data be leveraged to optimize business processes? Do you consider the Big Data challenges to be primarily technical or non-technical?

TK:
Business process optimizationThe opportunities presented by an ever expanding pipeline of information can be leveraged in seemingly countless scenarios. Whether observing web activity logs, social media activity, or call center recordings, the ability to synthesize vast amounts of data into insightful and actionable trends allows for a better answer to nearly any question related to business processes.

Though the challenges presented by the volume, frequency, and format of Big Data are varied, I see a greater obstacle in the technical barriers to implementing a big data framework. It is not a trivial task to capture and process such volumes of data from disparate sources in a connected and accurate manner. However, there fortunately exist many sufficient tools available to the ambitious data scientist to overcome both the technical and non-technical barriers of Big Data.

AR: Q4. What Big Data tools did you use for your projects? What data mining methods you found most useful - decision trees, naive bayes, SVM, neural nets, clustering, social network analysis, ...?

TK: The tools and modeling approaches used often vary quite materially from one project to the next. Some projects have required no more than SAS and coffee, while others have pushed us towards adoption of a Hadoop architecture with varied analytical software. In each case, the most useful data mining method has simply been the one to provide the strongest predictive strength. However, those techniques that are more palatable by the non-technical, such as decision trees and clustering, have often provided a good way to spur initial discussions in a big data project group.

AR: Q5. What do you personally think about the future of Big Data? Your predictions?

TK: I think Big Data, or more explicitly advanced analytics in general, will continue to play a more and more significant role in everyday business operations and decisions. Companies that fail to employ big data strategies will face a distinct disadvantage in today’s world. If the past few years are any indication, I would also expect the technology and methodologies surrounding data analytics to continue to improve and expand.

AR: Q6. What advice would you give to firms who are trying to make Big Data an integral part of their business strategy?

Strategy AdviceTK: It is important to approach Big Data problems from a somewhat academic perspective.

Companies should not be afraid to question their assumptions, acknowledge what they don’t know, and invest the time and energy into establishing long-lasting and impactful Big Data solutions. Be specific about what questions are trying to answer through Big Data analytics, and don’t wait until an analysis is complete to think about how it can be used to impact business strategies.


AR: Q7. You come from a solid background in Maths and Statistics. How important is it for Data Scientists to have Maths/Statistics understanding? What advice would you give to those building Maths/Statistics skills for a long-term career in Data Science?

TK: I think it is extremely important for Data Scientists to be proficient and knowledgeable in mathematics. Without a theoretical understanding of sophisticated analyses, one runs the risk of incorrectly interpreting results, building models with little longevity, and pursuing non-optimal methodologies. However, my advice would be to study these concepts in a very practical and applied manner, focusing equally on the computer science skills that are so often necessary to conduct and implement learnings from these analyses.

AR: Q8. On a personal note, we are curious to know what keeps you busy when you are away from work?

Private PilotTK: Despite assessing risk for a living and a moderate fear of heights, I’ve been working on a private pilot certificate for the last several months. When on the ground I am typically taking some sort of math or computer science course, tutoring a handful of students throughout the week, and seeking out sunshine and open water whenever possible on the weekends.

Related: