U of T Statistical Sciences: Tell us a little bit about your current work and projects.
Eric Cai: At Acosta, I use advanced methods in statistics and mathematics to evaluate the impact of retail coverage for manufacturers of consumer packaged goods (CPG). Retail coverage is a core service that Acosta provides; we send field associates to make sure that retail stores are ordering, stocking, and displaying our clients' products properly. This service costs money to the CPG manufacturer, so my job is to use statistics to measure the sales impact of this service and hopefully demonstrate that it offers a significant return on investment (ROI).
The three main components of my job are:
- methodological research and execution in measurement analytics
- business consultation for internal stakeholders and external clients
- cross-functional project management with product managers, database administrators, and technology providers
I use Python to prepare the data and execute a proprietary software that measures the impact of retail coverage on sales. I use KNIME to productionize my Python script into a point-and-click interface that non-programmers can use.
What are some specific areas of statistics that most interest you? What do you find fascinating about them?
I love Monte Carlo simulations. They are powerful tools for solving problems with no analytical solutions, they provide clean data to test new methodologies, and they are cost-effective to implement, both financially and computationally.
What makes you passionate about your job?
I am passionate about my job because I get to learn new skills all the time and do the type of work that I really enjoy. I have learned a lot about Python and the CPG business so far, and I'm also learning about the intricacies of implementing PySpark in Databricks. I love serving my internal clients; I communicate with them on a regular basis through emails and video calls. I get to combine my strong technical skills in statistics with my exceptional abilities in communication, project management, and B2B customer service. This is the first time that I have encountered graduate-level mathematics and statistics in a job, and I really enjoy the methodological rigour in our work.
When did you first become interested in statistics?
I became interested in statistics over ten years ago. I was considering a career in economics, and I took a statistics course to boost my credentials for entering graduate studies in economics. I enjoyed the course and found that I was good at it. At that point, a mathematics professor encouraged me to pursue applied mathematics, and a statistics professor encouraged me to pursue statistics. As I took more courses and learned about the career options of both fields, I concluded that the two fields had roughly equal potential for me within academia, but statistics had far more opportunities in employment outside of academia. Thus, I chose to pursue statistics.
Where did you get your undergraduate degree? What was your educational path like?
I obtained my undergraduate degree at Simon Fraser University. I finished with a major in chemistry and a minor in mathematics.
What made you choose U of T for your master's degree?
Being in Toronto was the biggest factor. I was close to many employers of statisticians and data scientists, and I could easily attend networking events to meet potential employers. The short duration (8 months) was very appealing, and the full academic funding was a decisive factor.
Can you share some particularly insightful moments from your time at U of T?
In my course on Statistical Computing (STA2102) with Dr Keith Knight, I worked on a project about Monte Carlo simulations of power functions. I enjoyed that project very much, and I earned a Travel Award to the 2012 Annual Meeting of the Statistical Society of Canada at the University of Guelph to present my findings. The audience praised my presentation very much, and I received a lot of positive feedback afterwards about it.
Were there particular courses that you found useful during your studies and/or after?
From my statistics courses at UToronto, the concepts that are most helpful to my career are about regression modelling with categorical predictors: reference coding, cell-means coding, effect coding, interactions, and interpreting the coefficients. Dr Jerry Brunner taught them in Applied Statistics 1. It turns out that categorical data are more common than continuous data in many industries.
Do you have any advice for current statistics students or those considering to pursue a degree in statistics?
When employers assess you as a job candidate, work experience matters more than your transcript. Make sure to get internships while you're still in school.
Networking is the best way to get jobs, and you should start doing that NOW. Aim to conduct one informational interview per week with someone who works in an industry of your interest.
Employers expect many skills that you may not learn in school, like Python, SQL, categorical data analysis, or natural language processing. Find out what these skills are in your industry of interest, and use your holidays and inter-semester breaks to learn them.
Learn logistic regression well. It is used far more than linear regression in many industries. Learn all the concepts about how to evaluate a binary classifier: sensitivity, specificity, precision, recall, receiver operating characteristic (ROC) curves, the area under the ROC curve (AUC or AUROC), positive predictive value, negative predictive value, and F-score.