Database Reference
In-Depth Information
Gutierrez:
Who finds this valuable?
Hu:
The entire music industry. The major labels are our biggest clients. There
is definitely a lot of value in providing these types of insights to the major
labels in terms of whom they should sign. We recently signed a record label
deal where we will provide specific artist recommendations and predictions
of who is going to be big in the next one to two years, and if they sign them,
we get percentage points on the record deal. Besides record labels, we also
work with promoters, band managers, and the artists themselves. So it is
really across the whole spectrum. I think it is all about being a thought leader
for the music industry.
For example, last year we published a post about our research on how social
media impacts sales. We ranked all of the networks based on how much of
an impact it actually has on your sales, and how you can predict future album
sales based on how well you are doing on any one of the different networks.
We found a rather surprising insight that Wikipedia has an amazing predictive
effect. We have to be careful not to say, “You need to drive people to your
Wikipedia page,” although that is somewhat true in the sense that Wikipedia
is a proxy for a deeper interest in you. So if somebody actually cares enough
about you to look up your background, what songs you have released in the
past, and who you have worked with—if she wants to know more about you
than just that one hit song that she's heard—then that person is much more
likely to buy your album. So that insight, I think, was very unexpected in the
music industry. And after we posted that, we have heard the research and
results cited all over the music industry.
Recently we also expanded into new verticals within the entertainment industry.
For instance, we have launched a book division and are now doing a similar
type of research around how social activity correlates to sales for publishing.
Gutierrez:
What was the first project you worked on after you joined Next
Big Sound?
Hu:
The first project I worked on was actually trying to clean up our data.
We have data from all different sources, from different social media sites and
different APIs. We have data that feeds in from our customers and from data
providers, and these are all very different sources that are being integrated
into the dashboard. The issue we run into is that we always have to deal with
missing and/or incorrect data. That is, I think, a problem that every company
deals with.
One of the first things I tried to do was to see if we could use machine
learning to predict what values were clearly wrong. One issue we found is
that there are a lot of cases where profiles are incorrectly connected in our
system. So we would see that Justin Bieber, for example, is the biggest art-
ist on Twitter as well as on a couple different networks, which is expected.
