You shouldn’t cut corners in cake baking or in data science.
That’s what we learned from our Chief Scientist, Jerimiah Hamon, at TVOT Live! recently.
During his Fireside Chat that keynoted the Data track, Jerimiah talked about how too much of any ingredient can ruin outcomes. That goes for adding salt to already-salted flour – a lesson he learned the hard way as a youngster in an Oklahoma kitchen – or for mapping what works when the content industry shapes decisions about what to deliver to audiences.
Jerimiah’s a master econometrician who has performed mathematical magic for an A-list of brands, including Disney, Amazon, Microsoft and more. With Firstlight Media, he’s spearheading our transformative approach to data aggregation and analysis as we help OTT streaming clients dive more deeply into the what, the how and the why of viewers’ content selections.
At TVOT Live! he outlined how we’re creating a cookbook of solutions that will help streaming providers zero in on each consumer’s entertainment taste buds – now and in the future. He talked about how our systems leverage the most-advanced cloud-native technologies to help our customers efficiently break data into manageable chunks, accelerating visibility that can be used to deliver “the opiate of the econometrician”– consumers watching each program they click on to completion. With the “backflip,” says Jerimiah, “we can go to the content provider and say here’s your library: first, you need A, B and C because people are searching for it and you don’t have it, and second, because they’ve watched all of this, we know they also want to watch D, E and F.”
Like culinary success, data mastery is a result of multiple factors. In Jerimiah’s case, it is deep knowledge of the underlying science; an understanding of human factors; and having the willingness and the patience to let go of biases and assumptions that prevent adoption of new methodologies. His advice:
- Take the time to get it right – A common mistake data scientists make is assuming that, just because an application runs and creates an answer to a given problem, it’s the correct solution. Instead, he stresses the need to prove that the answer is consistently correct, no matter how the problem is approached. “Validation,” Jerimiah says, “is 95% of the battle.”
- Balance computer science and human logic – Too often, data scientists don’t look at the data. Jerimiah points out that computer analysis could determine that he, Bill Gates, Elon Musk and Jeff Bezos are all billionaires, “but obviously that’s not true. If you aggregate a whale with minnows, then you’ll have minnows the size of dolphins. It doesn’t make sense, but a pure computer science method wouldn’t catch that.”
- Transcend your internal barriers –Sometimes you have to venture outside of your comfort zone – “take apples and oranges, mix it with steak and put coffee in the middle of it” – to produce the verifiable results you seek. “The barrier lives inside of us. As a species we do not want to see things.”
“You hear a lot of discussions around determining the genome for content,” Jerimiah says. “And it’s exactly the same thing as the cake analogy: you’re trying to find out which parts make a piece of content successful. Content is food for your mind; what we’re doing is building that ability to break content down into those very minute constituent parts and then learn from these aggregated populations that consume content.”
The rest is icing on the cake.