Though her path to the publishing industry may seem unconventional, Dr. Deepna Devkar’s journey from academia to the corporate world has always been on a data-driven trajectory. During her keynote at the Miami Publisher Forum on Mar. 11, “Experimenting in the Revenue Lab With Data Science,” the Head of Data Science for Dotdash will share key steps for injecting science into publisher data programs.
In 2014, Devkar and her husband uprooted their lives and moved to New York so she could do her post-doctoral work in neuroscience at NYU, one of the best programs in the country, to study the most fascinating computer in existence—the human brain.
“I was interested in memory,” she muses. “When we are bombarded with so much visual information, I wanted to know how the brain makes sense of it all and chooses what gets encoded into memory.”
Though she did not have access to a grant to study her true passion, functional MRI, she did have an advisor who had rhesus monkeys and a grant to study visual short-term memory. Rhesus monkeys are a great animal model to study the neural processes underpinning human cognition and behavior.
“I designed experiments, trained the monkeys to do short-term memory tasks to compare their performance with humans, and built computational models to test various theories of short-term memory,” Devkar says. “I have to say that I hated working with monkeys, but it was amazing to see how smart they are and how closely they resemble humans.”
“We are trained to think in a critical way. We write code. We use statistics to unravel what the data is trying to tell you,” Devkar says of data scientists.
And her training in scientific methodology was not lost on her new path. The same principles and analytical techniques drive her work today.
“Scientific methodology can and should be applied to all data,” Devkar says. “Whether it is understanding brain activity in response to a memory experiment, or understanding internet traffic patterns in response to seasonality—it ultimately boils down to time-series data analysis.”
“Studying human behavior in the real world is just as fascinating as studying it in the lab,” she adds. “It starts with asking the right questions.”
Human Learning Always Beats Machine Learning
Most people—even data scientists—wouldn’t think of data having a foundation in the publishing business, but it’s critical.
“Any time you need to ask where assets are being used most optimally, the answer always comes from the data,” Devkar says.
At Dotdash, she has built a team of Ph.D. data scientists who draw insights from data across each of Dotdash’s nine vertical brands. The most important part of analyzing your data is asking the right questions.
Devkar suggests looking at your own internal data as a starting point—Google Analytics, for example. Look for things like:
- What content is performing better than other content?
- Are there times of the day/week that yield higher traffic?
- Is some content more seasonal than other content?
- Who is your audience?
"Before building out capabilities for machine learning, start with human learning. The insights gained from your looking at your own data are far more powerful and pay huge dividends early on."
Once you see these patterns, try to determine the reasons for them.
“Before building out capabilities for machine learning, start with human learning,” Devkar says. “The insights gained from your looking at your own data are far more powerful and pay huge dividends early on.
“Innovate before you automate.”
Get to Know Your Audience
With more than 21 years of Internet data predating the existence of Google, nine premium brands and more than 90 million unique views monthly, Dotdash is doing something right when it comes to content and advertising.
And they continue to do it as they work to make their brands the go-to sites for specific content categories. For Dotdash, it’s all about making the best content on the fastest sites with the most respectful advertising.
Get even more insight into data science at Devkar’s keynote during the Publisher Forum in Miami, Mar. 10-13, 2019. Conference passes and hotel rooms are selling out fast!