ACM-W NA Profiles: Clara Yuan
Clara Yuan is a senior research science lead at Convoy Inc, a digital freight network startup. She holds a Ph.D. in economics from Virginia Tech. She has a bright intensity that she brings to everything from the world of transportation to her dog, Donut.
How did you come to data science?
My interest in data science honestly began with my transportation economics courses at MIT. I was lucky enough to take discrete choice modelling from Moshe Ben-Akiva, one of the leaders of the field. It was in that class that I learned to estimate a linear regression model for the first time.
Moshe used linear regression as a stepping stone to teaching us the more specific and sophisticated approach of discrete choice modelling.
Later on, I was looking for a change in my career. I found myself gravitating towards all the jobs with statistical modelling. I realized that statistical modelling and informing decisions in a rigorous, quantitative manner was what I really enjoyed. At the time, I didn’t think I had the credentials to land those jobs (which was probably not true), and ‘data science’ had barely been coined as a buzzword. I decided to go back to school for a PhD so that I could gain those credentials, and wound up writing all three chapters of my dissertation on discrete choice modelling. What goes around, comes around!
How does having an economics background change your perspective?
While a background in economics is certainly not unique in data science, it is in the minority. For most of my career, I’ve felt that my perspective is very different from common data science perspectives. That difference in perspective mostly presents itself in my focus on causal mechanisms: understanding how and why something happens, rather than whether and how often it will happen.
The analogy I like to use is selling umbrellas. As an umbrella retailer, you personally can’t affect the weather, but it heavily impacts your operations. On the other hand, you have direct control over the kinds of umbrellas you sell. Understanding why your umbrellas sell (for reasons besides weather) can help you strategize what to stock and promote.
Traditionally, economists have focused almost exclusively on ‘data science for humans’, because their results are used to guide policymakers (like estimating the benefit and causal mechanisms of education, healthcare or environmental protection to inform how much and which programs we should fund). The chief concern is on how policies will affect the outcomes of interest. I think many businesses have the same concern about how their actions and strategies will affect profit or volume, which is why A/B testing has become such a popular tool in the tech industry.
You are a person who loves stories and storytelling. How do you bring that into uncovering a narrative with data?
When it comes to interpreting results and providing insight, I’m not satisfied without a compelling story that explains the quantitative results and makes sense in the business context. Everybody on the team tells stories about data on a daily basis, because we’re always asked to explain what the data mean, why we got the results that we did. Unlike with most fiction, there isn’t just one, linear story, but rather a whole host of narratives.
As an economics-trained data scientist, my role is to generate those narratives and then find evidence for or against them. I’m lucky enough to work with a few other PhD-trained economists at Convoy. We frequently have lively debates over alternative interpretations of the results and identifying conditions to help us distinguish which narrative has more support. Deeply exploring the nuances of a complex problem space, accounting for those nuances rigorously, and arriving at a compelling story with robust evidence has made for the most satisfying moments of my career.
Convoy’s mission is to “transport the world with endless capacity and zero waste.” What has Covid looked like inside a shipping company? (Note: This answer is from mid-May)
First off, my views are my own, not Convoy’s, and I don’t speak for Convoy, only for myself. COVID-19 has certainly affected Convoy strongly, in a few different ways. Right around the time most of us were sent to work from home, Convoy experienced upsides from the rush on consumer products like toilet paper and disinfecting agents. As the economy slowed down and the full force of the pandemic really started hitting jobs and homes everywhere in the country, Convoy started experiencing the same challenges as the rest of the economy.
Now, truck costs are bottoming out, and a lot of carriers are finding it difficult to stay in business. Demand for shipping freight has fallen steeply amidst the economic slowdown, so freight is in a situation of oversupply. The tough economy is felt inside of Convoy, too – our team talks to carriers every day, and our hearts go out to them. Our product innovations like Automated Reloads, Power Only and our many-to-many model help carriers run more efficiently, which I hope will make a difference for carriers during this challenging period.
What advice would you give to a student or young professional?
This advice is what I would give to my younger self, even though she probably wouldn’t have listened to it. My advice boils down to one thing: don’t overplan. Some people are lucky enough to know exactly what they want, and the thing they want has an unambiguous path to achievement. My advice isn’t for those people, but rather for people like me, who don’t know what profession they want to pursue, or how to land the impact they long for. To this day, I still don’t know the answer to those questions, but, in the spirit of Edison, I know what hasn’t worked for me, and that’s overplanning.
When I was in college, I had every semester mapped out before I set foot on campus. Almost as soon as I had that spreadsheet saved, I had to throw it out and replan because I decided to change majors. As you can imagine, I repeated this process many times throughout my school years. After a particularly painful episode (dropping out of my MS program), I gave up on planning as a way of making life decisions.
Instead, I adopted a greedy algorithm: at every decision point, I choose the best immediate next step for myself, without ‘sacrificing’ my present happiness for some future goal which might never happen.
I was interning during grad school, I told my fellow interns, “I never want to become a manager. Never. I want to sit in a dark room by myself with nothing but the glow of numbers on my computer screen for company.”
Today, I’m a manager and I spend 3/4 of my time interfacing with other humans (it is through a computer screen these days, but I’m definitely not working by myself). People’s preferences change over time, especially as they’re exposed to different experiences, making it difficult to predict what will make you happy 20 or 10 or even 5 years down the road.
When you look at my career, it probably looks a lot more linear and intentional than it actually was.