Using behavioral targeting to customize content

30 Sep 2009 - 3:57am
5 years ago
9 replies
1143 reads
poomoo
2009

Hello,

At Scottish Television, we are actively exploring ways to serve up a
degree of custom content to every user the 'deeper' they travel
into our sites.

Taking a typical example of a user interested in football, we would
serve up more football related articles to that user. We would be
able to listen to the meta data on each article and page viewed in
the user’s journey, apply weightings and serve up more customised
pages.

As the user goes deeper into the site, we would narrow the focus of
articles and categories. As the user goes ‘back up’, we would widen
the focus of articles and categories.

The primary aim of "silent intelligence" is to create a continually
evolving and richer user experience through increasingly contextual
content.

We are making the assumption that a user wants more content on a
particular topic. As we learn more from what they are reading, we
want to provide ever more related content.

I am concerned that by continually 'narrowing the focus' of content
we could lock users into a dead end. However, it may not be any more
of a dead end than unrelated, uninteresting content. Although,
ultimately the user is the judge of this, regardless of the method we
use.

It's obviously important that we offer ways to 'widen the focus'
quickly. It's also important we balance the amount of content that
'narrows'.

I'd be really interested in hearing your thoughts on this, positive
or otherwise.

Thanks!

Comments

30 Sep 2009 - 12:08pm
poomoo
2009

Bumping back to top!

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Posted from the new ixda.org
http://www.ixda.org/discuss?post=46208

30 Sep 2009 - 12:37pm
.pauric
2006

As I interpret your description I believe you are building faceted
navigation
http://www.welie.com/patterns/showPattern.php?patternID=faceted-navigation

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Posted from the new ixda.org
http://www.ixda.org/discuss?post=46208

30 Sep 2009 - 2:26pm
Thomas Petersen
2008

We used to call the progressive contextual profiling. Did a big
project for Bank Of America back in 98 with that.

It makes sense to do this if you are trying to make sure that the
user get only what is relevant to them. (i.e. if you are from the
military, then information regarding military banking is important.

But it's very dangerous assume what the users want to see if you
don't have any clear idea about why they are selecting whatever
specific category.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Posted from the new ixda.org
http://www.ixda.org/discuss?post=46208

30 Sep 2009 - 3:22pm
poomoo
2009

Hi Thomas,

Yes, I've heard of banks using this before. It wouldn't be as harsh
as selecting a category to narrow the focus. We would require a user
to read several articles before we start to understand what they are
interested in.

So it would be a progressive, to steal your phrase.

It would be good to hear from others who have implemented similar
systems.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Posted from the new ixda.org
http://www.ixda.org/discuss?post=46208

30 Sep 2009 - 3:48pm
Thomas Petersen
2008

But isn't the problem that if you spend time understanding the
customer then you risk understanding what they used to be interested
in and not what they are currently interested in?

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Posted from the new ixda.org
http://www.ixda.org/discuss?post=46208

30 Sep 2009 - 5:42pm
John Gibbard
2008

I wrote about this in December last year
(http://ccgi.amorgos.plus.com/blog/?p=12) reflecting that as our
capabilities improve to track, trace and otherwise interrogate our
digital pathways we will increasingly be able to create post facto
hypotheses about what you are keen to see more of. A general
advancement of 10+ year old model of 'people who bought x also
bought y' model.

Care must be taken to not lose the value of serendipity in the
digital experience, particularly where compelling content such as
yours exists. Just because I read a lot about Scottish football
doesn't mean I will not be fascinated, even bowled over, at the
discovery of a classical concert being broadcast this weekend (excuse
me if I have interpretted your content incorrectly).

In addition to my piece already mentioned, there was an article in
August in the New York Times which addressed the decline of
'serendipity in the digital age'. I'd include the link but if I
nav away from this tab in Safari on the iPhone I'll lose the post
:)

I'm not suggesting that what you propose isn't right, I would just
counsel in favour of retaining some lateral discovery by
out-of-segment promotions and retention of the majority of the global
navigation architecture.

Best wishes,

John Gibbard
Senior Information Architect, Dare

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Posted from ixda.org (via iPhone)
http://www.ixda.org/discuss?post=46208

1 Oct 2009 - 6:39am
poomoo
2009

Thanks for your comments guys. Both of you make very interesting
points.

Thomas, I never considered this point of view; that the system would
always be one step behind the users current interests.

However, in reality, do users interests change rapidly? I would guess
a regular use exhibits the same habits over time, with subtle changes
in interest.

I like what you said, John, in your blog post,

"we need to ensure such personalisation enables us to exclude what
we know you will not like, keeping the edges of what you might like
very fuzzy indeed"

I love Last.fm too. I like the similar artist feature although, I
don't use it as often as I should. I also use Amazon recommendations
regularly.

I don't think it's fair to say these systems inhibit serendipity.
Users can still stumble across new things even if it's through a
system that learns what they like.

I think both your points raise important questions about the
algorithm and weighting in the system, rather than the principle.

To aid serendipity and discovery, the system could track trends
across the site, so we can see what the "crowd" is interested in.
These trends, our understanding of the individual user along with
editorial control could offer a more rounded experience of content.

Non?

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Posted from the new ixda.org
http://www.ixda.org/discuss?post=46208

1 Oct 2009 - 9:13am
Thomas Petersen
2008

With last FM the goal is simple. Providing you with suggestions based
on your musical neighbors. It look's at what you are listening to
and what you say you like or don't.

I can imagine that with you guys it's not that simple, unless you
where able to get statistics of what the users watch on their
television and whether they liked what they saw.

So I am wonder what you are trying to achieve.

Perhaps that is where we should really start?

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Posted from the new ixda.org
http://www.ixda.org/discuss?post=46208

1 Oct 2009 - 10:21am
poomoo
2009

Hi Thomas, our primary aim is to create a continually evolving and
richer user experience through increasingly contextual content. In
doing so, we would expect to see longer session times, increased page
views and increased ad revenue.

Our system, or the principal of Silent Intelligence isn't that much
different from the Last FM app. Our sites would continually listen to
our users consumption patterns:

* what they read across all our site
* what videos they watch
* what they choose to share

By understanding the user better, I'd hope we can provide and
environment that exposes more content they enjoy, more content that
is relevant to them and more content that they might not otherwise
have found.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Posted from the new ixda.org
http://www.ixda.org/discuss?post=46208

Syndicate content Get the feed