Interview: Carnegie Mellon Professor Ari Lightman On How Students Are Empowered By Learning To Use Brandwatch Consumer Research
By Kara FinnertyJun 10
Published September 3rd 2019
As a summer intern here at Brandwatch, I’ve spent a lot of my time learning about the data science team.
I recently managed to grab data scientist Katie Atwell for a conversation about her career path, her approach to working on tricky projects, and what an open company culture means to her.
Before Brandwatch, Katie had a long academic career in Psychology. She spent five years doing her PhD in a combination of applied social and experimental Psychology, using both qualitative and quantitative data analysis. While completing this she also taught research methods and statistics to undergraduate psychology students. She knows her stuff when it comes to data analysis and research.
After academia, she moved into a role as a technical data analyst at Brandwatch, where she got familiar with the workings behind the data. When the data science team was formed, she jumped into a new role as a data scientist and brings her psychology and data analysis experience to a cognitively diverse team.
Five years into Katie’s time at Brandwatch, I asked her if this background continues to give her a unique view when tackling problems.
Katie said that sometimes her background in psychology might inspire a project idea, but it’s mainly having a familiarity with problem solving and scientific method that really helps. “Brandwatch is dealing with data that comes from online, from humans,” she says, so her background in unpicking human data is really helpful for the team.
“Having a background in psychology means that you’ve had to think really hard about research methodology and experimental design.”
When I first asked Katie what a “typical day” was for her, she gave me a very concerned look. “If there is one?” I asked.
“There’s a huge variety to be honest – it depends what I’m doing,” Kate explained. “At the moment, for example, I have recently finished a couple of projects, one of which was building a sentiment classifier in Tagalog.”
Now that seems like an interesting topic. – developing an algorithm to identify the sentiment of text in a language spoken almost exclusively in the Philippines. That’s pretty impressive to me.
I wanted to know more about how Katie goes about getting project ideas to work on, and she explained that most of the time they’re inspired by either product or engineering. These can often lead her to new territory, too. “It’s gaining inspiration from other things that you’re working on, and wanting to take it in a different direction,” she says.
Katie explained that most of the team’s autonomy comes from the approaches they take to carrying out a project, which she finds is a really great way to work. “They might come to us with a problem but they’re not coming to us with what we should do as a solution,” she says.
I imagine that Katie often receives requests that just aren’t feasible, especially from people who aren’t as familiar with the data. As it turns out, the biggest problems start before even attempting to answer a question.
“Possibly one of the biggest struggles we have is just getting the problem defined in a way we can actually answer. A lot of communication needs to happen between data science and product to establish that.”
Right, so we have our question defined, what’s next?
Katie describes a very academic starting point, which involves “reading about whether anyone’s approached this problem before and what ways they tackled it.”
She then finds the data available and begins to work with it. “Really getting to know the data can take some time and a lot of what we do is cleaning data sets,” she says. Tedious as it may sound, this is an essential part of the process as it ensures the most accurate outcomes. “We then start with simple exploratory analysis before we start trying to tackle things, starting with the simplest approach first and then iterating to more complicated.”
Knowing that I’m a bit of a perfectionist, I asked Katie whether a project ever actually feels finished, or whether she is forever tweaking or changing it. Again, she says, it depends on the project.
“There is a certain amount of maintenance that comes along with any project,” she explains, “so you might find that you’re still answering questions or evaluating things years down the line.”
Being a marketing intern who’s had the opportunity to act on my interest in data science, I know first-hand the benefits of Brandwatch’s approachable atmosphere. Katie says:
“The relaxed atmosphere helps communication, there’s not really a huge hierarchical structure within the company so you can approach people no matter what level they are.”
It helps to bridge the gaps.
”You get to know your colleagues very well. You might have a discussion with an engineer that’s working on your project and come up with a really good idea to tackle the problem together.”
Along with her team, Katie handles vast amounts of data and provides insights to solve fascinating business problems. “You are continually learning,” she says, “and that can be exciting if you’re a data geek like me”.
Both being from science backgrounds, Katie and I bond over a shared love of research. She attributes the feeling of excitement when you get a significant finding as one of the main reasons she remained in research.
“I am learning something new most days, which is incredible for someone who is addicted to learning like me.”
Olivia Swain interviews VP Data Science Hamish Morgan on the inner workings of Brandwatch’s "coolest" team.