Database Reference
In-Depth Information
Gutierrez: How did you end up at MailChimp?
Foreman: I met with Neil Bainton, MailChimp's COO at the time, and
spoke with him about the problems at MailChimp and how data could poten-
tially solve some of them. Because it's a successful online company with two
domains in the Alexa 500, the data set being produced is just massive. So it
was an amazing opportunity to use data to solve problems. I initially thought
in my meeting with him that we were talking because he was looking for tal-
ent, so I started suggesting other people he should speak with. After some
conversations, he asked, “Well, what about you?” As we talked more about it,
the scope of the opportunity became clear to me, and I ended up here.
It's been an absolute blast, because MailChimp operates so differently from the
government or large enterprise companies. In the government, you have layers
and layers of management, and there are really well-defined chains of com-
munication. You find the same thing in the enterprise world, too. MailChimp
is a small organization, so it's very flat. It's not the kind of place where you
go and tell other people what to do or go talk to their manager and their
manager tells them what to do. No, it's all done through close communication
and collaboration, and so it was a new challenge to do analytics in this type of
environment. I have found that I really liked the independence and the chaos
that came with it.
Gutierrez: When did you realize the power of data?
Foreman: When you look at operations research problems or optimization
problems, the data's important, but what's really important is the formulation.
You have a really small amount of input data—and really complex decisions
need to be made from that data. When I went and started doing price optimi-
zation models, that's where the power of the data really hit me and opened my
eyes. This is because you're forecasting demand for some product, whether
it's a hotel room, or a room on a cruise ship, or a seat on an airplane, or some-
thing else using historical data, and then updating the pricing of this product
as more data comes in.
What's amazing about these pricing models—I think they're some of the coolest
data science models around—is that their decisions directly affect revenue. If
I choose to lower prices on my hotel because I think that's going to maximize
revenue, that's a pretty audacious decision for a model to make. The fact that
it can say, “You need to take the price down $15,” and have that actually affect
your bottom line in a positive way is amazing. The model can actually say that,
and then we can go back later and prove it was right. We saw some Fortune
500s get 2 percent revenue uplift per year, which is amazing at that scale.
That really opened my eyes to the fact that if you can appropriately gather
and keep track of lots and lots of transactional data, there's a competitive
edge there.
 
Search WWH ::




Custom Search