In today's content-rich world, consumers are overwhelmed by the
choice available to them on TV, Video on Demand, etc. Initially
this is exciting for the new consumer, but it rapidly becomes an
irritating experience as, each time, they click through channel-after-channel
on the EPG or wander through the VOD library.
The result - people stick to watching what they already know, and
after a while begin to question whether they are getting value for
their subscription money. In this situation customer retention becomes
a problem, and raising ARPU similarly challenging. The fact is consumers
cannot select what they don't know - so 'Search' is the wrong approach
to a constantly moving EPG or large VOD library. Intelligent, personalized
'Recommendations' solves this problem, thereby improving retention
and raising ARPU.
Essential Taste Broadening
To improve customer retention and grow ARPU, it is essential
to expose customers to a wider range of content than they would
normally discover for themselves.
Where the customer explicitly declares a wide range of tastes this
is relatively simple to achieve, but in most cases (especially during
the critical early phase of the relationship with the recommendation
system) the customer is cautious/lazy and will declare a narrow
range of tastes or even none. For instance, he will often start
by declaring only gender, age and perhaps one taste area. In response,
instead of recommending a default Top 10 list based on collaborative
filters (which leaves the customer repeatedly stuck with the same
recommendations until he trusts the system enough to start giving
feedback), ThinkAnalytics system generates a fresh and diverse list
of programs every time, using feature-based sub-genres linked to
the declared profile. In this way, different programs are selected
from the live EPG, including 'Long Tail' surprises or one-offs to
intrigue the customer, and as a result he is encouraged to divulge
more and progress his relationship through this crucial start-up
phase. Without this sophisticated approach, even customers declaring
a wide range of tastes risk receiving recommendations which are
no more than a 'reminder service', whereas ThinkAnalytics ensures
the customer is regularly offered broadened content.
Seeding the Application using Local Data
It is critical that a customer's first experience is a good
one. We therefore do not use data based on other companies or
countries because all markets and TV providers are different in
terms of content and culture. The whole point about 1-2-1 personalization
is to provide a localized, differentiated customer experience
from day one, otherwise customer's will stop using the service.
You need a Dutch system for Holland, a French system for France,
etc, all differentiated to your customer base. Because ThinkAnalytics
technology trains itself, the seed can either be created from
a set of sub-genre template associations that your Marketing Department
has defined, or ThinkAnalytics can automatically generate associations
from focus group data that you have collected. Either way, this
seed is rapidly enhanced by customers providing real feedback
ratings.
Automatic Classification of your Content
All your content is automatically classified by the system.
This means:
- No ongoing reliance on external manual classification/editorial
coverage services
- No ongoing fees for external manual classification services
- Syndicated programs can be automatically classified in a
consistent way
- One-off programs are easily classified
- TV series can be classified by episode where appropriate
- You retain editorial control of the automated classifications.
Business Rules control marketing bias
This ensures the system can balance strong consumer centric objectives
with those of the business's marketing objectives.
Keeping Recommendations Fresh
Many successful retail recommendation systems are often criticized
for having too long a memory which causes it to make out-of-date
suggestions. ThinkAnalytics avoids offering suggestions based
on old data by applying ageing factors. In addition, it understands
time-of-day, weekday v weekend, and seasonal variations; and being
a real time system, it responds instantly to customer feedback
and to the constantly evolving content.
Who owns the data?
The data collected is hugely valuable as it reflects not just
genre preferences but individual program likes/dislikes. This
ongoing data and derived intelligence is owned and controlled
by you, available only to you, stored wherever you choose and
accessible in any way you wish. You therefore control your ability
to differentiate yourself - a key requirement in the commercial
world.
Open Application
ThinkAnalytics' technology uses a range of open, industry standard
algorithms with which any commercial analytics person will be
familiar. Coupled with an open, component based architecture,
this ensures you are:
- in control of the future direction of your recommendations
application
- able to select alternate analytic algorithms when necessary
(thereby avoiding the limitations of a one-size fits all, black-box
solution)
- able to extend the use of important consumer learnings to
any part of your organization such as marketing, retentions,
package configuration, acquisition, channel marketing, etc
- able to ensure the application meets your corporate style
requirements and fits the local cultural preferences of your
country and your customer base
- able to integrate the application into all your customer
touch-points.
Focus on: Content Discovery
