How it Works





The HoneyLizer transforms users’ extensive social data
(Facebook likes, interests, events, friends, posts,
etc. – often 100 attributes) and demographics (even
behavioral attributes, when available) into unique
social data signatures. These social data signatures
are clustered and are highly efficient predictors of
in-app conversion behavior.
The game designer uses the HoneyLizer UI to select the
event elements and options that the HoneyLizer will
optimize in real-time.

The entire Integration process is rapid, and normally
takes 1-2 days (depending on number of app events
selected for optimization).

Once integrated, the HoneyLizer rapidly creates the ultimate conversion
matrix by methodically serving different page options to users with
different social signatures. When a conversion occurs the HoneyLizer
adds that signature to that event’s conversion cluster. The process is
repeated until sufficient conversion data is positively correlated. This
learning period can take a few hours or up to a few days depending on
conversions volume.
The HoneyLizer is now ready to optimize live incoming traffic, and
immediately jump-improve the designated conversions. Even first-time
users get immediately optimized – no user history is required!

Once the boot-strapping period is completed, the
HoneyLizer automatically engages its proprietary
predictive algorithms to automatically serve users, in
real-time, the page options they will like better (and
convert better!).
As it optimizes, the continually improves its
predictive accuracy, to better reflect current users’
preference patterns.
The HoneyLizer often finds compelling yet unintuitive
correlations that no analyst can ever discover, thus
also providing deep new insights into users preferences.

Consider the following example of a game on Facebook:
In a treasure hunt game users can be presented with one of two payment page options - paying 50 cents to
purchase an energy boost and getting a 10% bonus, or paying two dollars for a much larger energy boost and
getting 30% bonus.
Several actionable correlations can be discovered; for example the HoneyLizer may discover that users with the
attribute of less than 50 friends and like books convert better when presented with the 50 cents default
option, while users with more than 300 friends that are “night owls” convert better when presented with the
larger energy boost option. No A/B testing scheme can ever achieve this advantageous result! The HoneyLizer
automatically invokes the best conversion page option for each player resulting in higher conversions.

© 2012 Bees and Pollen. All rights reserved