“When I tell someone I work at Hopper, I often get an excited reaction, and they tell me a story about how much money they saved on their last flight with us.”
Those are the words of Hopper Data Scientist Matt DiNardo, and that excitement for the company’s work seems to permeate the rest of the data science team at Hopper — even if the analytical among us don’t comprehend all the data behind it.
The company’s mobile app uses machine learning to provide travelers with the best possible flight based on their needs, and it takes a lot of data to do that. Fortunately, the data science team makes that highly technical information more palatable for the non-data savvy among us and acts as the foundation of the company’s overall data-driven culture.
We caught up with three members of that team to learn more about what they do and how they help define a culture so pivotal to the company’s overall mission.
EMPLOYEES: 300; 62 locally
WHAT THEY DO: Hopper is a mobile platform that travelers can use to book their flights and hotels. It uses large datasets and machine learning algorithms to make pricing predictions and give users recommendations into the best places to go and the best times to book.
WHERE THEY DO IT: Boston
IDEAL CANDIDATE: Potential Hoppers interested in joining the data science team should have communication skills as sharp as their technical skills. The team helps shape the product roadmap for the company, so having the ability to turn data into stories and actionable insights, both internally and externally, is important.
Ella Alkalay, Vice President of Data Science
Ella is in charge of leading the company’s data science team. Her previous experiences working at Outbrain, IBM, and in the mission control centers for the Israeli Air Force and Israel Aerospace Industries have given her a wealth of experiences she has used to triple the size of Hopper’s data team and build out the organization’s data-focused culture.
DUSTY TRAILS, CLEAR MINDS: Ella takes backpacking very seriously. She spent nine months hiking from Ushuaia, Argentina — the southernmost point in South America — through 10 different countries before ending her trip in Mexico. The activity allows her to free her mind and return to work refreshed with a new sense of clarity.
It’s been a few months since we last checked in with you. What is your team up to these days?
We have more than 40 million users now, so we have to experiment wisely while continuing to accelerate algorithms in production. We need to establish processes and build tools that allow us to continue to experiment and deploy algorithms to production in short sprints without negatively impacting users or breaking the app. We collaborate closely with our engineering and product teams to bridge those two things. A major project we worked on is a machine learning platform that allows data science to train, evaluate, deploy and update algorithms in production in a self-serve way.
In addition to using data science to bring transparency and guidance to our users as they’re planning their travel, data science is key to every decision made at Hopper. So, our team is collaborating with key stakeholders across various departments and lanes to ensure that we’re answering the right questions and guiding the product roadmap. These days, we are working on improving our self-service user analytics platforms to get closer to our customers by scaling data access and visual insights internally.
Everyone gets to own their projects and feel empowered to make decisions.”
You’re leading a team of data scientists at Hopper, so how would you describe your leadership or management style?
Our team is very autonomous and independent. Therefore, everyone gets to own their projects and feel empowered to make decisions. I’m there to provide guidance and support at any time. I foster an environment of knowledge sharing, collaboration and open conversation. We’re a close-knit team, and I love getting to know everyone on the team on a personal level. This week, for example, I hosted the team at my place to treat them to a traditional Israeli dinner.
You have quite the resume — Outbrain and IBM, among others. How is Hopper’s company’s culture different than previous companies in which you’ve worked?
I’m always amazed to see the fast turnaround time between ideation and production. Hopper moves very fast with a strong bias for action, which allows us to test new ideas and grow quickly. The work environment encourages customer obsession and cross-team collaboration, which leads to some very creative thinking.
We have a highly data-driven culture, and the data science team has a significant role in determining the company’s roadmap. To work in a company where data is the product’s key value proposition ensures that the data science team has the greatest impact, and the company is constantly tackling fascinating data-driven challenges.
Hayley Berg, Economist
Hayley frequently collaborates with Hopper’s public relations and business development teams to make the company’s data-driven insights digestible for the less-technical parties it engages with: consumers, the media and airline partners.
KEEPING THINGS MOVING: Whether horseback riding, hiking, or sailing all around New England, Hayley loves to be outside doing any activity she loves. Spending time logged off and soaking in the fresh air helps her to refocus and feel rejuvenated to tackle the next week’s projects.
You’ve been here more than half a year now. What was the largest obstacle you faced during your onboarding and how did your team help you overcome it?
When I joined this past winter, I didn’t code in the programming languages we use at Hopper, so there was a steep learning curve from the beginning. All new team members are paired up with a mentor to accelerate their onboarding, which helps with everything from integrating socially to learning the ins and outs of our data. Outside of my direct mentor, the team was super supportive from the beginning, carving out time to teach me the nuances of our databases and helping to answer questions as I got up to speed on the new languages.
We ask ourselves what a leisure traveler would need to know about a particular topic and how they would want to absorb the information.”
What’s a recent challenge you experienced in translating data to be more digestible, and how did you overcome it?
It’s not uncommon for media outlets like NPR and NBC Nightly News to pick up our research and run segments quoting our analysis. When this happens, it can be a challenge to cut a full length, data-heavy analysis down to a 10-15 second consumer-friendly sound bite.
To make sure we’re always putting out consumer-friendly research, we start every analysis from the perspective of the consumer. We ask ourselves what a leisure traveler would need to know about a particular topic and how they would want to absorb the information, and then develop the analysis from there. Since we’re starting analyses from the mindset of the everyday traveler, it’s now simple to translate those insights at the drop of the hat.
What’s something about your role or your team that may surprise people?
Having worked and studied in highly analytical fields for most of my career, it was not uncommon for me to find myself working among teams that were almost entirely male. At Hopper, our data science team — currently 50 percent female — is led by our awesome VP of Data Science Ella Alkalay. It’s empowering to be surrounded by such a strong cohort of highly technical women leading our product and team forward.
Matt DiNardo, Data Scientist
Matt’s work as a data scientist allows him to drive impact in a number of different areas of the business. He works on Hopper’s price prediction algorithm that informs users of the ideal time to book their flights, while also working on the hotel-focused side of the house, where his team runs competitive analysis for travelers to ensure they’re getting the best prices.
PLAYING THE KEYS TO SUCCESS: Matt has found pairing the piano with his work is like pairing a fine wine with a great meal. He keeps a piano next to his desk at home and takes breaks to practice and decompress while he’s running a Python script on a large dataset or training a model.
What tools does the data science team currently use?
Given the size of the datasets we work with, we use PySpark for large-scale distributed computing. When working with smaller datasets, we use Python with the pandas’ library. In terms of data visualization, we have a variety of tools we use depending on what we need — Tableau for daily reporting, Plotly for interactive graphs and some R here and there for exploratory analysis.
We have no shortage of data, and we have no shortage of data problems — a great combination for a data scientist.”
How does your work at Hopper keep your professional passions stoked?
We have no shortage of data, and we have no shortage of data problems — a great combination for a data scientist. Our data problems cross so many different domains in data science — from price intelligence to recommendation systems and geospatial analysis — that you are always learning and applying new techniques.
What are some of the aspects of your team that make it unique?
“Another day, another dataset,” is a motto that rings true for us because we are always running so many experiments to find ways to provide more value to our users. Every day brings more evidence that allows us to prove the worth of good ideas and learn from bad ideas.