Validate Efforts Through Experiments | Measurement Fundamentals | App Marketing | Udacity
- Articles, Blog

Validate Efforts Through Experiments | Measurement Fundamentals | App Marketing | Udacity


By now you know how important it is
to create goals and look at metrics. But other than tracking those goals,
how do you affect them? Your routine, as a marketeer at a startup should
be based on trial and error. Experimentation and iteration is
the most important thing for a startup. Let’s hear what some of our
experts have to say about this.>>We did some sort of
behavioral studies of what a new user to YouTube,
how they utilize the service. And it typically, what we would see would be the first time they would come
visit this site would be through some kind of viral marketing that may be they
saw one of their friends share a video. Maybe it was on Myspace, maybe it
was an email, check out this video, this was really funny. But the first time generally it
doesn’t really stick with them, they don’t remember this site,
maybe they’ll remember the site. They will remember this site, but they won’t remember the site that it
came from, because I think we get that all of the time where people would
just be sharing the instant messages through emails, just check this thing
out on the Internet, it’s very funny and you see it, you laugh and
then you close the browser window. But I think the second time, there
would be another time where somebody would share this, another video. Maybe it’d be somebody else sharing and
then after a few times visiting the site, it starts kind of, the logo
starts kind of thinking in a bit, the sort of domain names start
sinking in a little bit. And what we would see is that people, they themselves would type in
YouTube.com into the browser and they would actually go
to the site themselves. And then from there, It’s about
sort of uploading their own videos, finding their own videos and sharing it,
so it kind of spreads out that way. But we’ve always kind of looked at it
required a lot of sort of effort with the design of the site to be able to say
within the first five seconds, that’s the amount of time window that you have
to be able to really grab the users. And so there was a lot of work that was
done on the related videos on the right side, because the video that you were
sent may not be interesting at all, or may not be up your sleeve in terms
of what you’d be interested in, but something on the right
side would grab you. Same thing on the front page, there
would be ten videos, so we’re always any page on YouTube there was always one
link away from starting another video. And so in case you got bored,
instead of closing your browser, I think psychologically it’s always easier to
kind of spend more time on a site, rather than closing the browser and
going to do work. And so there’s always something
that’s within one click that can start another video and
if you don’t like that, again, then there’s another,
one click away from another video. One significant part of YouTube is not the first video
viewed but the subsequent video. That whole experience and so
whether, and we’ve different metrics giving us different times, but
it has been as high as something as, or it has been as low as sort of
the average length session of YouTube is like 12 minutes, but it’s been as high
by some metrics as 28 minutes and that’s incredible when you consider
the average length of the videos. Whatever, 30 seconds to minutes and
so well, one important part of that is and
not the first video. We’re assuming that people
are already getting to the site but how do you get them to watch and
click on the second video and then after that how do you get
them to click on another video. And more importantly cause like if
they get disappointed by that second, third video how do you still get them
to click on the fourth, fifth, sixth, seventh video. And the solution to that,
partly, is the related videos on the right side and we always put that
high because that was going to be your channel to get into more content,
but then the follow-up question. And this is where we had the smartest
engineers at YouTube working on this day end, day out, which was how do you
populate that related to video? One idea of thinking about
the content on YouTube, is you can think about it and sort of almost
different islands that are connected. Somehow, but different islands, they’re
islands about sort of singers and the different types of
genres of artists out there. And then there’s islands of sort of
movies, there’s islands that related to fashion and food and there’s islands
that’s more directed towards kids. So there’s all these different islands,
some of them are bigger, some of them are continents and
some of them are tiny. But people are very
passionate about that and they’re all interconnected with graphs
and if you can visualize that then it’s kind of what if you are watching
a single video on one of these. Maybe it’s a big continent and
maybe it’s a small island. What do you populate with all
the other videos on the right side? And the obvious approach is if you’re
watching a content about cooking, everybody on the right side
is more about cooking and so it’s all within the same island. But what if by the fifth video they’re
just sick of watching cooking videos, how do you get them to
watch something else? And so we realized early on, but there
was a lot of testing involved with this, is, what if we actually start stuffing
in different types of content in there. So what would be coming back from
search results would be the things that are most related with this video is
obvious it’s going to be cooking. But let’s just say artificially we
take out 50%, 75% of this content. We just make it so it’s impossible
to show more than 25% related videos is cooking, then you start
actually feeding in this other content. Maybe we find out that people that are
interested in cooking also like soccer. And so we start actually putting in 10% of the videos which
are related to soccer. And started playing around with this and
so maybe people that are interested
in this abstraction of soccer and sports if they’re watching sports
videos maybe they’re also interested in watching football or maybe they’re
interested in watching tennis. And so we started playing
around with this idea, what percentage is tightly related
to the video that you’re watching. How much of it is, it’s not unrelated
ever, but there was a algorithm that went behind, how do you actually
create this sort of loose relationship? They’re still related, because we do
know that people that watch this content also watch this content. But it would be hard to explain if you
didn’t look at what the actual watching sessions and patterns are to be able to create that connection between
say food and football. Well, that’s probably
an easy one to create, but so we started doing a lot of
research along these lines and trying to create algorithms to be able
to better guess as to people that are watching this video would also be
interested in watching this video. But it took a lot of theories and
testing those theories, and pushing stuff out live and
seeing what the users would actually do, and tweaking and
tweaking and tweaking that.

About Ralph Robinson

Read All Posts By Ralph Robinson

Leave a Reply

Your email address will not be published. Required fields are marked *