Egret Endeavors is the founder of Attribute Profiling (patent pending), the next generation of collaborative filtering.  Why is it the next generation?  Take a look at our White Papers or read below to learn more.

"Why" Attribute Profiling

Let's take an example. Let’s assume you are going to recommend two products.  One product to a dear friend and another to a complete stranger.  Why is it so much easier to accurately recommend a product to your dear friend rather than the stranger?  Of course the answer to this is obvious – you know you’re friend’s likes and dislikes whereas you don’t know the stranger at all.  This simple example highlights the essence of Attribute Profiling and its eloquent and common-sense approach to collaborative filtering.  That is, Attribute Profiling captures “why” someone likes or dislikes a product while, at the same time, seamlessly allows you to compare those preferences to products for an accurate recommendation.

Your users shouldn't be strangers.

Stars vs. Tags vs. Attribute Profiling

While there seem to be many flavors of active collaborative filtering the current state of the industry can be broken down into either "Stars", the ubiquitous five-star rating paradigm, and Tags, users creating one or more keywords to attach to a product. Let's take a quick look at each and then see how Attribute Profiling fits in.

Too fixed

Stars are a very structured and comprised of a very closed set of input data for the user. Structured in the sense that the user has only a limited choice of values to choose from. It is akin to a multiple-choice question on a test. The user simply has to think about a few options and not worry about anything else which, inherently, frees the user from exerting any unnecessary effort. This is good. I believe people like structure just for this reason. We are all busy and if we can get the results we want with less energy then more power to us. This structured simplicity is why the five-star rating paradigm has been the most successful and widely distributed form of active collaborative filtering.

However, it does have it's limitation -- the very closed set of data input for the user. By this I mean the user can only choose between a limited, numerical value. What if the user wanted to say more? What if, for a movie review, he/she wanted to say "the acting was terrible but the cinematography was amazing!" You'll never capture that with Stars. The key thing about Stars, however, is that it captures how the user feels about a product (but not why he/she feels that way).

Too open

Tags have no structure and a completely open set of input data for the user. It simply solves the limited input data problem by completely opening it up. The user is free to choose whatever they wish. It is akin to an essay style question on a test -- absolutely zero structure and takes lots of energy. While Tags obviously fill a niche that Stars do not the biggest disadvantage is that it doesn't capture how a user feels about a product at all. Tags simply describe a product to help categorize and identify similar products but does absolutely nothing to link that information back to how the user feels about that product (let alone why).

Just right

Attribute Profiling hits the "sweet spot" for collaborative filtering. Its unique approach allows for the best of both worlds -- and then some. The beauty of Attribute Profiling is that it provides, in one eloquent solution:

• Structure for easy data input,
• A pre-defined set of input values that can be customized (a subset of open values), and
• The ability to not only capture how the user feels about a product, but, most importantly, why he/she feels that way.

View our white papers to learn more or, once you’ve decided to more forward, see how we can support your implementation.