multichannel connect
careers
all access

Technology

Cover Story: What’s On?

5/11/2010 3:16 PM Eastern

Talk about a one-sided affair.

Each month, Americans
spend upward of 150 hours
gazing longingly at their TVs.
Yet most TV services, surprisingly,
don’t provide any feedback to their
human companions. They just sit
there passively, obediently changing
channels on command.

Until now. Pay TV services
are starting to get smarter about
suggesting movies and shows to
viewers. Dish Network, AT&T and
Verizon Communications are
among those pursuing new automated
recommendations today,
and cable operators are actively
exploring the technology as well.

For the most part, they’re following
the lead of other services
— TiVo, Netflix, Apple’s iTunes
Store and Amazon.com, to name
a few — which have been recommending
video selections based
on someone’s preferences or order
history for years.

TV providers are banking on
the fact that an intuitive guide,
like a great concierge, can induce
you to watch more TV. And
that can affect business in several
ways: by cutting churn, differentiating
services, driving up
revenue from video-on-demand
purchases and upselling programming
tiers.

The need for more personalized
recommendations is evident,
considering VOD libraries
are already approaching 20,000
titles and could grow five times
as large. Add to that the prospect
of millions of Internet video clips
accessible on TV, and it’s obvious
subscribers need a new tool
to find content.

“If you think about it, for someone
to say, ‘Gee, there’s nothing
on TV tonight,’ is kind of nonsensical,”
said Mark Hess, Comcast’s
senior vice president of advanced
business and technology development.
“Especially on the VOD
side, as we increase the number of
titles, recommendations become
more and more important.”

As simple as the concept is —
recommending a title — the execution
can be surprisingly difficult.
For starters, set-top capabilities are
limited. Meanwhile, several people
in a single household often use the
same account, diluting the effectiveness
of a personalized suggestion.
“The recommendation might
not really reflect what the person
who’s sitting in front of the TV
wants,” Hess pointed out.

Privacy issues may trump recommendations.
Some MSOs,
under their existing subscriber
agreements, may not have the ability
to monitor a viewer’s habits and
then provide recommendations.

Dish Recommends featureAnd some viewers say the
idea of a company tracking their
choices — and saving them — is,
well, creepy. When TiVo first introduced
its recommendation
service, the running joke was “My
TiVo thinks I’m gay.”

Nonetheless, pay TV operators
need to do something on this
front to keep TV services relevant
and valuable. “We get stuck in this
‘paradox of choice’ when we’re
presented with massive numbers
of options,” said NDS Americas
sales director Paul Ranger.

By the end of 2010, several
cable operators will have
implemented recommendations
with VOD the first target,
predicted Corey Ferengul,
Rovi executive vice president of
product management and marketing.
“One of the big ways of
driving more VOD is telling people
what’s available,” he said.

Content recommendations fall
into four main forms: based on
content (“if you liked Avatar, you’ll
like these shows”); based on popularity
(“here are the top TV shows
being watched right now”); based
on user profile or behavior (more
tricky — see the aforementioned
TiVo gaydar); or through social
networks (e.g., “most of your
friends love The Closer”).

Several technology companies
are pitching solutions that
make smarter TV recommendations
to viewers, including Rovi,
Jinni, ChoiceStream and ThinkAnalytics.

Each offers a different take. Jinni
— whose content-discovery system
was voted “best product idea”
at the CableLabs Winter Conference
2010 in February — groups
content by genre, which is one of
the most typical ways to organize
content. But the startup’s system
goes well beyond that, providing
additional linkages and recommendations
based on the emotional
attributes of a movie or TV series
(like “mind bending,” “gloomy” or
“race against time”).

“You can create for yourself a
semantic genome that shows you
exactly what you like,” Jinni CEO
Mike Pohl said.

Initially, most operators are
opening the door on content-based
recommendations and those based
on aggregated-viewing metrics.

Comcast is kicking around the
idea of introducing an interactive
TV application,
based on
CableLabs’
Enhanced TV Binary
Interchange Format specification,
which would display the
top 10 shows being viewed at the
time or suggest content related to
the TV show or movie a subscriber
is watching. “I think it’s better
to have the wisdom of the crowds
rather than trying to predict what
an individual viewer wants,” Hess
said.

Dish Network in April debuted
a recommendation feature with
the launch of its ViP 922 Sling-Loaded digital video recorder.

The system uses viewing data
from Dish’s 14 million subscribers
to determine four related TV
shows and movies, airing within
the next nine days, and provides
recording options.
“The business goal is definitely
to get a stickier service, to give
customers an easy method to find
the shows they want,” said Keith
Gerhards, director of software engineering
at EchoStar Technologies,
which supplies Dish with
set-tops and related software.

It took time to refine the Dish
recommendation filters, which
are updated once per day. EchoStar also found that the system
needed to be contextually aware
— so as not to make an inappropriate
suggestion. As Gerhards
put it, “We had to make sure that
only G-rated content gets recommended
with G-rated content.”

The “Dish Recommends” feature
uses a filtering algorithm
developed by Cambridge, Mass.-
based ChoiceStream.

AT&T’s U-verse TV also employs
ChoiceStream for its VOD “top
picks” feature, based on a subscriber’s
past video-rental history.
For example, if you’ve rented comedies
or children’s movies in the
past, the app will recommend the
latest titles in those categories.

The telco’s app also provides a
list of the top 10 on-demand titles
U-verse TV customers are renting
and the ability to rate a movie you
have previously rented, which is
averaged into other U-verse TV
customer ratings and displayed
next to the title in the on-demand
recommendations list.

VOD recommendations encourage
consumers to buy more
on demand and are effective at
converting non-VOD or free-VOD
subs into paying customers, said
ChoiceStream chief technology
officer Mike Strickman. In the
first year after recommendation
features are deployed, an operator
should expect VOD revenue
and usage increase around 10%,
according to Strickman.

“We have really
tried to focus on
the transactional applications
generally because
the revenue you’re generating
is measurable,” he said. “It’s
harder to tell how much impact
you’re having on the subscriber-retention
front.”

AT&T also uses ChoiceStream
to deliver personalized upgrade
offers. The telco’s promos select
from around 50 shows—which
are available only in a higher-priced
programming package—
that an individual subscriber
would likely be interested in given
his or her past viewing history,
such as Showtime’s The Tudors or
HBO’s True Blood. Strickman said
the personalized offers are up to
five times more effective than generic
upsell offers.

Verizon, for its part, offers
“Recommendations for You,”
which provides personalized VOD
recommendations for customers
based on programs they have
previously viewed. Its “More Like
This” feature recommends VOD
titles that are similar to others recently
selected by a customer, and
a “What’s Hot on FiOS TV” widget
lists the most popular programs
and VOD titles currently being accessed
in a viewer’s ZIP code.

But there are stumbling blocks
for many cable operators. Set-top
box platforms may be one of the
biggest gating factors.

“There’s no place in the interface
today where recommendations
exist,” said Brian Kahn,
director of engineering for VOD
vendor SeaChange International.
“It tends to take a long time to
redesign the guide, and get tested
and deployed.”

Changing the interface in a set-top
guide is a much bigger project
than adding a new feature to a Web
site, Strickman said. “Web development
is pretty mature, whereas
these set-top stacks are much more
proprietary, more complex, running
on devices that have very limited
resources,” he said. “And politically,
there are all the decisions about
what goes on the set-top.”

In addition, if an operator
were to start recommending
content from a linear channel,
that could create conflict with
programmers.”Convincing NBC or
TNT to make suggestions based on
their programming to drive viewership
to another channel is going
to be tough,” ActiveVideo Networks
senior vice president of marketing
Edgar Villalpando said.

And privacy looms large on
the content-recommendation
front.
For the most
part, providing recommendation
features
based on personal data can be delivered
on an opt-in basis and fall
within accepted use, said David
Jacobs, chief technology officer of
Amdocs' broadband, cable and satellite
division “Most subscribers do
trust their service provider.”

Another tough problem for TV
content recommendation is that
television programming is harder
to analyze than movies. Let’s say
you religiously follow the Chicago
Cubs. That doesn’t mean you
would necessarily have any interest
in, say, watching a Mariners-
Royals game.

“Just because I like sports
doesn’t mean I should record
every sports program,” Ferengul
said. “Movies have kind of a
known criteria.”

Questions abound: Should
TV recommendations
cover only primetime?
How are news programs recommended—
should potential political
slant be factored in? Is The
Daily Show With Jon Stewart
a
news show or a comedy show?

What’s also difficult about
TV is that often the newest content
is the most relevant. But if
it’s brand-new, it won’t have any
viewing metrics yet, Strickman noted.
To address this, ChoiceStream’s
recommendation algorithm takes
into consideration metadata such
as actor, genre and release date
for a piece of content that has not
been rated.

Perhaps the most vexing challenge
is how to handle multiple
viewers in a household with
different likes and dislikes. This
week, Cox Communications
takes the wraps off Trio, its new
interactive program guide that,
among other features, lets up to
eight different family members
set up their own preferences. The initial
release of Trio, however, doesn’t
include any content-recommendation
features.

The multiuser-rating problem
has been around for years. TiVo, on
its customer-support site, provides
this notice: “If some TiVo Suggestions
do not reflect your preferences,
it is possible that other members
of your household are pressing
‘Thumbs Up’ on shows you do not
care for, and vice versa.”

Even Netflix, which runs the
granddaddy of video-recommendation
engines with a database of
more than 3 billion ratings, hasn’t
cracked the code on accounting for
different tastes of people in a single
household. “Anything that comes
highly rated with Adam Sandler, I
know it’s a rating from my teenage
daughter,” Netflix vice president of
corporate communications Steve
Swasey acknowledged.

It’s important to keep in mind
that recommendations — while
increasingly vital — are one of several
tools to promote content. In
Demand Networks is using multiple
forms of search, navigation and
promotion to drive up VOD usage,
said CEO Bob Benya. Those include
In Demand’s “online storefronts”
with up to 20,000 titles available
for download-to-own or rent; using
Web portals to let subscribers create
VOD playlists; promoting VOD
assets using a pop-up prompt on a
linear channel; and integrating on-demand
into linear search.

At some point, smarter suggestions
will find their way to the set-top,
Benya said: “Personalization
and recommendation tools are
very powerful.”

VID-PICKERS
Key video-recommendation players:
ChoiceStream (choicestream.com): Powers recommendation
features of AT&T U-verse TV and Dish Network

Jinni (jinni.com): Content-discovery system uses “moods” and
emotions to recommend movies and TV shows

Rovi (rovi.com): Acquired content-recommendation firm
MediaUnbound in March 2010

ThinkAnalytics (thinkanalytics.com): Uses real-time analytics
and behavioral-modeling techniques applied to linear TV, video-on-demand, games and music

TV Genius (tvgenius.net): London firm’s customers include Sky,
AOL, ITV.com

SOURCE: Multichannel News research

September