Christopher earned his MS in Statistics in the Summer of 2012. He currently works for LinkedIn's Marketing Solutions Team as a Sr. Data Scientist. Here's what he had to say about his experience in our program!
- Q: Please describe your current job title/company and what your day-to-day is like.
- A: I am a Senior Data Scientist working on the Marketing Solutions team at LinkedIn. LinkedIn, as a place for professionals in the workforce to make connections to others, serves ads on its platform. The Marketing Solutions team monitors how efficient the ad serving experience is for advertisers. Are people on LinkedIn clicking their ads? Are they purchasing the products advertisers are selling? Is the process of creating ads on LinkedIn a seamless one, or are their parts of LinkedIn's experience that advertisers struggle with? As a Data Scientist, I use statistics to answer these questions and highlight areas of improvement.
My day-to-day is about 80% of the stuff that comes before you can create that fancy machine learning model: aligning with other teams that a proposed analysis will actual answer the questions they want it to answer and what data we need to get the answers we want. This usually takes the form of collaborative Google Documents that I work on with Product, Engineering, Design, and User Experience teams. After this, I'll conduct due diligence on the data to make sure it is instrumented the way we expect it to. Early on in my career, I learned that "garbage in, garbage out" is one of the most important lessons you can learn working with real-world data!
The actual modeling is only a small part of my day-to-day, but an important one. To ensure we are conducting rigorous analyses at LinkedIn, we have peer review sessions where we share our work and are critiqued and questioned by our peers. The sessions are intense and thought-provoking, and dramatically strengthen the quality of the work!
Q: What made you choose our graduate program?
A: I did my undergraduate at Davis and fell in love with the quality and intimacy of the Statistics program. I had a feeling that an undergraduate degree in statistics was only barely scratching the surface of what was out there, and I felt like the department had much more to offer me if I pursued the graduate program. Despite having a lot of undergraduate students in Statistics, there was a level of personalization I experienced in my upper division courses that strongly appealed to me. Whereas my lower division and biology classes were auditoriums of hundreds of students, my upper division classes would sometimes be less than 40 students. I usually sat in the front row, and the stats professors had so much personal experience to share about working on real data problems and how statistics could be useful. I loved their perspectives and wanted to go even deeper with graduate study.
Q: What was your first job after graduating with your MS from UC Davis?
A: I worked as a biostatistician in the Pharmacy Benefits division at UnitedHealth Group. We ran these reports called clinical programs, where we would look at millions of people's medical and pharmacy insurance claims and determine if they were not using medication correctly, or if they could save money by pursuing other treatment options. As a biostatistician, my job was to comb through claims data and identify which people were eligible for intervention from our team of pharmacists, who would reach out to members or their doctors and inform them if changes should be made to their treatments. I also did large scale observational analyses where we would look at the effectiveness of treatments in practical real-world use. Often times, drugs that looked effective in a clinical trial being administered in a hospital would not be effective out in the real world when individuals had to take medication on strict schedules, or where the drugs had side effects that reduced patient adherence. Most of my day was spent writing SAS programs.
Q: Can you describe your career path from your first job after UC Davis to your current position?
- A: I was writing programs in SAS at UnitedHealth, and because we were working with patient data, our analytical techniques were usually quite conservative. Machine learning was becoming a pretty hot topic, and I wanted the chance to do more modeling work on interesting datasets that would find my work producing actual changes to how the company operated. From there, I moved to San Francisco and pursued a career doing data science in the tech startup world. My first job in tech was a social network where one person would do karaoke to half a song, and the platform would invite other people to sing along and then mix the two parts together to form a duet. It was a pretty cool idea, and my work was frequently used to shape the vision of the product, based on what features resonated with users, what new features users wanted the most, etc. My boss from that job brought me along to a toy robotics company where I did similar work, but where the analyses were much more open-ended. We started the Analytics team at the company, so there was no precedent at the company for any of the work we did, which was truly exciting. It really felt like a supportive and collaborative environment, and I had a blast learning how to work on projects with non-statistical audiences when we needed to define what core metrics we cared about (how do you quantify "fun"?). That company shut down, and I reached out to a close colleague of mine who got me into the LinkedIn Data Science pipeline, where my skill set working on performance marketing analytics in an autonomous manner made me a good fit for the team. I've been at LinkedIn since August of last year.
- Q: How did your program and experience at UC Davis prepare you for where you are today?
- A: I took a wide variety of classes at UC Davis and worked with a lot of real-world messy data outside of classes. Both are crucial to being an effective data scientist in the real world.
Seeing a diverse variety of statistical techniques prepared me to recognize them in my own work. For example working on survival analysis problems in biostatistics proved essential in estimating customer lifetime value, and I used material from a seminar I attended as a grad student about Bayesian Time Series to create a solution for detecting unexpected changes in revenue. I didn't realize it at the time, but building up a corpus of examples of how statistics could solve problems allowed me to solve my own problems, with some modification.
Effectively dealing with and presenting messy real-world data, a skill I acquired after taking Prof. Temple Lang's Statistical Computing class, made me a far more efficient data scientist. You can't get to the flashy data modeling and prediction part if your data isn't in a form that your modeling functions require. Even worse, if you haven't confirmed that the data makes sense and is correct, nothing else you do matters. The world's best machine learning model can't solve for Engineering forgetting to record when a customer makes a purchase on the website!
- Q: What were the most influential people, courses, and/or experiences for you at UC Davis? Why were they so significant?
- A: It would be too hard to name all the people that had an influence on me at UC Davis. The top names that come to my mind that sparked my passion for Statistics were George Roussas, who taught my first upper division statistics class and helped me along considerably as an undergraduate, Frank Samaniego and Christiana Drake, who started me on the path of working in the messy world of applied statistics, and Paul Baines and Duncan Temple Lang, who sparked my interest in statistical computing.
I have used material from each class I've taken at least once in my career so far, but if I had to give credit to just one class in particular, it would without a doubt be Professor Temple Lang's statistical computing class (STA141, at the time). It was extremely difficult, but no other class has come anywhere near close to matching the overwhelming usefulness of the subject material, and I owe >90% of my proficiency in R to him.
Lastly, the most enjoyable and memorable part of my time in the UC Davis Statistics program was the time I spent with my grad school colleagues, especially my cohort and my office mates. A lot of life events occurred while I was in graduate school, and without their support inside and outside of class, I think I might've been lost to follow-up instead of surviving!
Q: What advice do you have for students interested in a career in Statistics?
- A: There's no single unique job that fits a Statistics degree. The current trend in Statistics-educated individuals seems to be Data Scientist, but the title is often a Rorschach Test on which enthusiastic employers foist all their hopes and dreams for "what the data will tell them". The job is rarely that magical. Be specific about why you want to pursue the field. A "machine learning engineer" has a very different day-to-day and skill set than a "data analyst", "data scientist", "associate professor of statistics", or any of the dozens of Statistics-adjacent jobs out there.
The number one thing I get asked when prospective data scientists reach out to me via email and LinkedIn is what skills employers look for in new data scientists. I tell them that in my four jobs since I left academia, I look for three things (in this order):
- Acquire their own data using some flavor of SQL.
- Solid intuition and explanations of the foundational statistical tools (what is a confidence interval, what good is a p-value).
- Clear analytical framework when solving a problem (why is this question worth answering, how much impact will it have, what metrics should we measure and why).
Q: Anything else you'd like to share with us?
- A: As a student, I vastly underestimated how valuable humility and amicability were to being an effective part of a team and creating a great place to work. If you are pleasant to work with and aware of your own weaknesses and shortcomings, other people will want to be around you, and you'll have more fun at work, too.
Finally, if you have any further questions, or wanted a more detailed description of an average day-to-day from a small sample of data scientists (n=1), feel free to reach out via LinkedIn (linkedin.com/in/christopheraden).