Database Reference
In-Depth Information
Smallwood:
A great source is the amazing number of articles out there
these days. Just do a Google search for entertainment optimization, or pre-
dictive models for entertainment, or related searches, and a lot of interesting
things will pop up. I would also say that talking to people, of course, is a great
way to do it. Go on LinkedIn and find people who are working in those kinds
of companies and reach out to them. Most people are happy to have a half-
hour phone call, which can really give you a lot of information as well.
Gutierrez:
What in your career are you most proud of?
Smallwood:
I would say of my whole career I'm most proud of the team that
I lead right now. I've hired a bunch of great people and the fact that we're such
a tight team and super fun to work with is incredibly gratifying to me.
If you were looking for more of an individual output thing, I'm probably most
proud of some work I did at Intuit prior to it acquiring Mint. We had a
scrappy little team of four people doing an internal startup-like project. I had
the chance to lead the creation of a personalization system. It was Mint-like
in that we were using a recommendation engine to match a couple hundred
advertisers we signed up and who had coupons to people based on people's
spending behaviors. It was super exciting to build a whole recommendation
system from scratch that actually worked quite well. It contributed to Intuit's
decision to acquire Mint, because the project was sort of a proof of concept
that we could do it and make it work.
Gutierrez:
What is a typical Netflix day for you and your team?
Smallwood:
It would be quite different for me versus my team, so I'll talk
about my team. Across the team we do a collection of things, like the strategy
meetings where we debate ideas and output. We also do that internally as a
team. For example, we have a brainstorming meeting every week with a spe-
cific agenda for everybody who works on experimentation. Somebody comes
with a topic to present and discuss, and then at the meeting we try to figure
out things like, “What should we do about this problem?” or, “I had this great
idea. Do you guys think it's great? I think it's great. Or does it suck?” We do
that internally as a team, as well as with different parts of the organization.
Another chunk of time is spent on sharing results, ideas, and data with other
teams where there are dependencies across the organization.
A fair chunk of time is on the source data. We always have more things we
want out of the data, so we work closely with other teams who either log
data, or do the data warehousing, or do the business logic around the data.
Working with those teams is always a big part of our team's work.
We also spend time studying model results and iterating on models. This
includes the modeling piece as well as the tactical piece of whether a model is
working. And if it's not working, what are some of the ideas that are around
that we could try differently? So studying results here as well.
