Database Reference
In-Depth Information
people feel like it's more spacious and so they pack in more appropriately.
That's not an analytical, data-driven solution, rather it's more of a design solu-
tion. And that's kind of the perspective you get when working with all these
other kinds of people.
Gutierrez:
Was math combined with beautiful design already present in your
work or is it a new thing since you joined MailChimp?
Foreman:
No, it's something that has developed since I joined MailChimp.
I worked at Booz Allen for many years working as a consultant for the
government. Then after that I was at a boutique consulting firm called Revenue
Analytics, which builds large-scale pricing models for Fortune 500 compa-
nies. These consulting engagements offered fascinating opportunities for data
scientists. The problems were complex, affected top line revenue, and
had excellent data sets. You can make the argument that companies like
Intercontinental Hotels were doing big data before there was big data. But
given the customer on the receiving end of these models, and given their already
ugly, difficult-to-use set of enterprise BI tools—I won't name names!—what
I provided was rigorous but often less-than-beautiful.
When a customer topics a hotel room or a seat on an airline, they don't inter-
face directly with a pricing model, so who cares if it's ugly or not? I worked
with one Fortune 500 company on a very complex production optimization
model that involved a lot of decisions, a lot of money, and a lot of raw materi-
als. The model was a huge cost-saver for their business and even garnered
some press coverage. It was all true. None the less, the “user interface” was
an Excel sheet hot-glued to an Oracle database, IBM CPLEX, and some other
systems. Don't let the words “Fortune 500” fool you; the big guys can do work
to match anything janky the start-up world might produce.
Getting back to my current job, one of the cool things about working at
MailChimp is that I get to build analytics products that may be used by the
customer directly. One thing I've worked on at MailChimp is Discover Similar
Subscribers. For this project, I built a clustering algorithm that helps users
detect segments within their email list. But just as much thought went into
displaying the tool to the user as went into the math.
I had to work with others to figure out how to communicate data mining con-
cepts to the layman in a way that they could use and understand the product
effectively. So now I became not only a data scientist/analyst, but I also became
a communicator/translator to a broad group of people. And that's where this
interplay of math, beauty, and design comes in. How do you think of data sci-
ence in a user experience context to make your data products really work
for businesses? It's new to me at this job compared to previous analytics jobs.
It's been refreshing.
