At The Trade Desk, data drives everything. It’s how the digital advertising platform targets audiences for client ad campaigns, which means every facet of the organization relies on data scientists to fuel its efforts.
We spoke with a few members of the data science team to learn how they work with data, how they work cross-functionally, and how technology helps them innovate.
FOUNDED: 2009
EMPLOYEES: 1,000+; 330 locally
WHAT THEY DO: The Trade Desk is a technology platform that empowers brands and media buyers to create, manage and optimize data-driven digital advertising campaigns across all formats and devices at scale.
WHERE THEY DO IT: New York, with offices in Bellevue, Boulder, Chicago, Denver, Hamburg, Hong Kong, Irvine, Jakarta, London, Los Angeles, Madrid, New York, Paris, San Francisco, San Jose, Seattle, Seoul, Shanghai, Singapore, Sydney, Tokyo, Toronto and Ventura.
NOTABLE PERKS: Employees and their dependents receive fully covered health benefits and receive a new laptop every two years — which means they can keep their old one.
LUNCH ROULETTE: Every month, a group of cross-functional employees gets together and has lunch outside of the office, on the company dollar.
SALES SUCCESS: From the way an account manager works in tandem with a trader and sales lead to build and maintain client relationships to each team’s holistic approach to giving and receiving feedback, collaboration is everything at The Trade Desk.
Mark Davenport, GM of Data Science
Mark runs the global data science team that is responsible for developing products and deriving insights from the massive amount of data that The Trade Desk platform generates. His team then works closely with the product and engineering teams to see this data reach the client-facing platform.
BEYOND WORK: Mark loves running. He ran the Nashville Marathon in April and the Brooklyn Half Marathon in May, and he's on the hunt for new races in the fall.
Your background is in finance. What made you decide to pivot to data science for a digital advertising company?
Right out of undergrad, I joined an investment advising company that specialized in asset-liability investing. At the time, the title of “data scientist” wasn't really a thing, so that wasn’t my title, but the responsibilities were the same: developing proprietary models, engineering features, building scrapers to generate data sets for research, constructing automated tools to democratize data around the organization, and more. I actually loved the work and the company, but I wasn’t excited about the finance industry.
Around that time, The Wall Street Journal ran a series of articles about this new side of the advertising industry that was leveraging massive amounts of data to serve better ads through real-time bidding. I grew fascinated by it and started doing more research into digital advertising and the types of jobs that existed. I made the decision to go back to grad school to further my skills, and now here I am.
Tell us about a project that this team produced under your leadership that you’re proud of.
Last year, we had the largest product release in our history. The product, called the Next Wave, also included the release of our suite of artificial intelligence tools called Koa. Koa was the culmination of nearly two years worth of effort by the team and included contributions from every single data scientist at one point or another. One financial publication even called the Next Wave our “iPhone moment.” It was an awesome moment for the company generally and for the data science team specifically. It gave us a true sense of identity within the broader engineering organization.
The big thing for us this year is hiring and successfully scaling out the wonderful team we have today.”
What’s next for data science at The Trade Desk? How is your team moving the company forward?
The big thing for us this year is hiring and successfully scaling out the wonderful team we have today. Another major focus is to help our clients buy better ads through our platform. We support that by building best-in-class optimization algorithms that help clients pay the right price for the right ad to the right user. We’re also developing effective ad measurement tools to help clients move beyond last-touch attribution methodologies toward causal measures and multi-touch attribution models.
Yi Fang Chen, Senior Data Scientist
Yi Fang is responsible for building different statistical and machine learning models, prototyping them and working with engineers to bring them to production.
BEYOND WORK: Yi Fang spends most of her time outside of work talking — and bargaining — with her two toddlers.
Talk to us about the problems you solve when you come into work and the technologies you're using or building to accomplish those goals.
I am building a statistical model to identify features that will help us decide whether an audience has viewed an ad from a particular advertisement on linear TV. This helps us increment the reach of audiences to bring in extra revenue to our Connected TV (CTV) platform. To achieve this, we utilize the data from different resources, including some automatic content-recognition data from Samba. I am using PySpark and R and run the data on AWS EMR cluster.
What role does data play in shaping your company’s product or technology?
Data drives everything. We need the data to better serve and recommend advertisements to clients. For example, we need to study data for the profile of anonymous users to help us determine whether we should recommend an ad to an end user or if the user has been pre-exposed to the ad. Every product decision relies on analysis that we made based on data.
Data drives everything. We need the data to better serve and recommend advertisements to clients.”
How does the company culture foster an environment for tech innovation?
We have an applied AI lab, which fosters creation, research and collaboration across different teams. It enables data scientists to not only focus on day-to-day production work but also provides the freedom to explore and research guild projects utilizing modern AI machine learning technology. If the guild project is successful, we can pitch the project and also bring it to production.
Karen Schettlinger, Senior Data Scientist
Karen develops mechanisms and finds optimal methods, processes and algorithms to automatically identify audiences that advertising clients can target to effectively run their campaigns.
BEYOND WORK: Along with reading and traveling to foreign countries, Karen enjoys crafting things by hand, especially via crocheting, knitting, stitching and origami.
What solutions are you working on at The Trade Desk, and what technology is helping you?
Our company processes up to 9 million ad requests per second. That’s a lot of data. Data science is extracting the relevant pieces of information and then finding or developing the right methodologies to optimize an advertising campaign to meet our clients’ goals. This includes finding the right audience that a campaign should be targeting, determining when and where to deliver the campaign, or deciding on the optimal price to bid for each ad request. We work closely with the engineering department to implement our models using technologies such as Spark, Scala, Python and R.
How does your team make the data digestible for your company and users? How does data influence your company?
Most of the work we do happens in the back end. We automate our processes as much as possible so that a user of our platform can pick an audience they want to target with a simple drag-and-drop or tick a checkbox to optimize their campaign delivery and enable cross-device targeting or something similar. We also work with other teams such as business intelligence, engineering or UX to provide reporting and monitoring in a form that is easily digestible for both internal and external clients using our platform.
Depending on the product or project, there are likely more teams involved somewhere along the process. Everything we do is a team effort.”
How is your team working together to evolve The Trade Desk’s product?
Our team works in close collaboration with a lot of other teams to evolve The Trade Desk’s offerings. For example, a request for a new product may come in from product management, so we’ll work with data engineering to get to the relevant data in the right format, then perform our research and model development and work further with engineering to productize our models. Finally, we work with our traders to check usability and real-time performance. Depending on the product or project, there are likely more teams involved somewhere along the process. Everything we do is a team effort.
To stay in the loop with everyone, we use Slack for chatting and Zoom for video calls. You’ll always see people from other offices around the world on your or your neighbor’s screen. That helps bridge the gap between long distances and makes it feel like we’re all in a gigantic global office together.
