Database Reference
In-Depth Information
people in that team are working on, but at least they have a primary focus
so that they can have all the domain knowledge that goes with that part of
the business.
Obviously the marketing organization has totally different strategies that
they're thinking about and totally different kinds of data they're looking at than
the product organization. So we've found that it's really important to have
people on our team who completely understand that part of the business.
Otherwise, how would you do that translation in your head of what you're
trying to model? This partial specialization also helps us know how much
load is coming in from the different areas, and then we are able to readjust as
needed on an ongoing basis.
Ongoing, there are things that pop up that you didn't expect, and suddenly you
feel like you have fifty percent more work to do as a team and wonder how
you are going to do it. Since I pretty much understand most of what's going
on at Netflix, and folks within the teams understand their areas as well, we're
pretty good at figuring out what to de-prioritize in order to let the super-fire-
drill thing come in the door.
Gutierrez:
How do you know you're solving the right problem?
Smallwood:
I think it really comes down to understanding the business goals
and understanding what the priority of the business is for the thing that you're
trying to solve. Is it a curiosity, or is it something that's actually going to be
used for an important decision, or for an important process or product? Or
is it something that's operational in nature? And is there already something
there that's pretty good and you're just looking to optimize it? Or is it some-
thing where there's nothing's there, and it's just a glaring issue where you
know you can improve things dramatically?
There's a judgment call there that comes from understanding the business
priorities, as well as understanding what you know you could offer from a
modeling or algorithm standpoint. I do think it has to do with understanding
the data at hand, what's possible with the data and models, as well as what's
important from a business perspective.
Gutierrez:
How do you know you're capturing the right data?
Smallwood:
That is a really fascinating question for me, because I feel like
through the process of my career at the different companies I've worked with,
I've seen it all. At Yahoo!, all data under the sun was captured and it was great.
When you wanted to study something, you knew you could find the data
somewhere. However, it was challenging because the data was a mess, which
meant that the ratio of the time you spent massaging the data and getting it to
the point where you could actually use it for something, compared with the
stuff you're actually doing—was high.
