As you know, I’m interested in solutions that help consumers figure out what to buy. This holiday season seems to be a breakout one for gift recommendation services like Gifts.com, Target’s Gift Finder, Best Buy’s Gift Ideas, and Walmart’s Top Gift Lists. When talking with Yahoo! Shopping a little while back, I found out that the company’s Gift Finder service is partly powered by ChoiceStream. I followed up with Darren Gill, ChoiceStream’s VP of Business Development, about a month ago. Here’s a brief introduction to the company.
What is ChoiceStream?
“ChoiceStream is a personalization solution provider. We deliver the right selection to each user on a one to one basis.”
Many people know of personalized recommendation services from Amazon or Netflix? How are you different than those services?
“Most of those services are based on collaborative filtering which is powerful, but it’s agnostic to what the actual content is – you could replace the content with numbers and make predictions on which numbers are right based on an overlap of the content.”
“For instance, you buy a blender, vacuum cleaner bags, and The Last Samurai. I buy a blender, coffee table, and vacuum cleaner bags. [The collaborative filtering system] might say that I will like The Last Samurai.”
“With our system, we have attributes – based on the preferences in the space – on all content. If we look at movies/music – we want to think about the music as dark, aggressive, or the movie as intelligent, silly, or funny. When a user rates content [like movies and music], he is talking about the underlying attributes. We look at the attributes of The Last Samurai that are consistent with other movies. It’s a content domain expertise approach. There are two halves – understanding of content attributes and understanding how users approach content. Without both, you can’t do one-to-one [personalization]. The real power is that people are unique and with personalization, your uniqueness can thrive. We want to be able to push more of the content that’s useful to the user.”
[A little more...from a realease about ChoiceStream working with eMusic: "Instead of focusing on traditional collaborative filtering, clickstream or activity analysis techniques, which are notoriously unreliable predictors of behavior and require extremely large volumes of data in order to make meaningful predictions, ChoiceStream focuses on understanding the underlying attributes of the content being personalized as well as each user's preferences for those attributes. Armed with this insight, ChoiceStream matches each individual with the content best suited to his particular tastes and preferences."]
Is e-commerce a minor focus? You also have solutions for mobile, satellite, and cable?
“E-commerce is really important for 2006 – we want to take what we’ve done at Yahoo! Shopping and go out and reach some verticals.”
What technology are you actually providing with Gift Finder? Where do you start and Yahoo! Shopping takes over? How does your technology work? It looks like Yahoo! Shopping is just grouping/tagging products – 25-34yr old appropriate, female appropriate, relationship appropriate, etc.
“[Tagging] is a great starting point. But let’s talk about scaling that. How many resources does it take? Now people are going in, and you’re giving them products and categories and you need a system to figure out what’s working. We built a system that tags it automatically; the build creates the tag. We learn, we look at the clickstream and the tag, who is clicking through on the path and what are the attributes of those products.
The more people who do that, the more we can bring product [recommendations] into the path.”
“[With ChoiceStream], the merchandisers can be in a much more fun role where they tweak things as opposed to the really heavy effort [of tagging/categorizing].”
“We’re grouping products around the attributes as opposed to just looking at the products. At Yahoo! Gift Finder, it’s not currently one-to-one. It’s gathering data.”
“Tomorrow it becomes your personalization shopping experience. As we see what you bought, we can go one-to-one with you. That’s when it gets fun and that’s the future. The next step is to use all that to seed our models so shopping search with personalization will give you different results than someone else. You put in digital camera and [based on personalization] we might give you tiny ELPHs vs. SLR Digital Cameras for $5000 with billions of pixels.”
Comparison engines do a great job of telling you where to buy and how much to spend for a product, but they still don’t answer the question what should I buy? Is ChoiceStream the answer? Is social shopping – where you shop alongside your friends – the answer?
“I think these are many options. With eMusic, we spent tons of time building up great features. And we wanted people to be engaged. Choicestream is additive. All users have unique interests and different things will speak to them about different products. “
“On the community side – sites like Epinions, ezboard, and Trip Advisor are a great way to get people one step closer to buying, but the people have no idea of how much credibility to give to a reviewer. For example, with Trip Advisor, someone might give a hotel a 10 on a scale of 1 to 10, but that person is only a 1 on a scale of 1 to 10 in being similar to you.”
“We’ve built a product with eMusic, we call it like minded recommendions, and take people in the community and compare them based on attributes. There can be powerful discovery through a community. Music you like, movies you like, but it applies to shopping too.”
