[MUSIC] Social media's just really recently
taken off as a marketing channel. So what we're seeing is that sort of on
surveys among CMOs, they say, right now, social media's about 10%
of marketing budgets. In 2018, it's projected to be 20%. We're seeing that increasing amounts of
marketing dollars are getting shifted from more traditional channels
towards social media more broadly. So this would involve both
advertising through social media, as well as firm trying to can a foster
consumers to talk about their brands by providing platform reviews,
by sometimes responding to them and does it a lot of ways which try to do it. So seem that channel and if you ask CMO is
about 64% of CMO's is think word of mouth, online word of mouth is one of the most
powerful marketing channels in the future.
But at the same time,
it's actually not that well understood how quantitatively strong these effects can
be, and that's where our study comes in. What we try to look for is, if you will,
an external shock to the system, something that changed how much people
were talking about a particular product or brand, that wasn't related to any
underlying characteristics of the brand. So if we wanted to avoid us that more
people are talking about a brand because more people bought it in the past.
So essentially the talking about
is a result of the sales, but we want to establish is that whether
talking can actually lead to more future sales in that realm. And that's usually hard to do. [MUSIC] We look at Sina Weibo which is the Chinese
equivalent of Twitter if you will. A very similar type of platform. Twitter itself is blocked in
Mainland China so it cannot be used. So Sina Weibo emerged as a substitute for
that. And what happened is that in 2012,
following a political scandal, Sina Weibo was partially blocked for
3 days and so what we're going to investigating
is we're going to look at TV shows. A lot of them have
a very active following.
A lot of them are aired at a daily level,
so we see every day people talk about these
episodes then they watch it at night. Then suddenly for three days lot of this
activity is gone due to this external influence, which is the censorship event
that's coming from the government. And so essentially looking at that
something happened to show us that tended to have a lot of following and suddenly
disappears during those three days, do we see that being reflected
in their viewership numbers? So the censorship event helped us a lot in
pinning something down that's otherwise very hard to do. So most studies prior to that. Usually it's something like
relating sales of a product to the number of Tweets pertaining to the
product, maybe from the previous day or week, something to assure that's
not happening at the same time.
But even then, it's hard to know
whether that's the right attribution. So you might imagine again, the example
that might be one high-quality product which constantly has a lot of sales and
constantly has a lot of Tweets about it. Anther product does not have that, and even if we look at temporal differences
between sales and the microblogging that precedes that, that might still give us
just a correlation rather than actual and impact of the microblogging,
in a causal sense, onto the sales numbers. And there's a lot of reasons to
think that the, if you will, more naive approaches tend
to overstate the effect. So the incorrect measurement would tend
to go towards overstatement rather than understatement of the effect,
the reason being that probably the most sales reason is that
the actual cause of a relationship could be from sales to word of mouth
rather than the other way around. And this might be misattributed. So it could be the case that
you sell the product a lot and because a lot of people bought it they
then go on social media and talk about it.
So it's really the sales that
precede the word of mouth rather than the other way around
which is what we want to establish. Now in our case, we can rule that out
because the censorship event wasn't based on sales numbers or anything,
it's really external to the system and its impacts micro blogging activity and we can then trace through
the effect of that on to sales.
And we can be sure that that
reduction microblogging does not originate from sales differences. Which would be the case if we just compare
different products with each other. [MUSIC] So what we find is that relative
to the typical movement in TV ratings doubling the number,
doubling micro blogger activity moves it by about
20% of the typical movement. It happens from episode to episode,
which coming from just one advertising channel in this realm
is actually a fairly significant. The findings are significant in two ways. So one is, I think both intuitively for a lot of practitioners as well as in
the prior literature, academic literature, we often thought these effects might
be even larger than what we find.
So to the extent that firms
are increasingly shifting their marketing dollars into those channels it's very
useful to know that maybe these effects might have been overstated and we have
to be careful in sort of managing them. Secondly it's a realm where it's just
incredibly hard to study these effects because user dragged content isn't
under your control as a firm. So you can't manipulate it in order
to experiment around with it.
So you actually have to resort to
methods like the one that we're using, which we call quasi-experimental. We're not doing the experiment ourselves, but we have something that is
almost like an experiment. The Chinese government,
if you will, experimented for us by shutting it down for three days. They didn't mean to shut it down for
TV shows, but it so happened that it's something that We can leverage to
actually understand the affect.
It's a study that cautions
a little bit against overstating. That is sort of the magic bullet
in terms of marketing channels. So it doesn't seem like it's clearly
out performing other channels. So it seems to me if anything we still
want to stick with a certain mix of marketing tools, rather,
diverting everything into that realm. [MUSIC].