WEEK-02 Resource - Loomia, a recommender system packaged as a web service !!!

A web developer decides to deploy a recommender system to provide a more personalized experience to those who visit his web site and in the process promote more business by exposing more of his product line to his clientele. His options? Develop a recommender system from scratch, use code from open source projects, or depend on third party developers to develop a system for him. Potentially expensive options. Loomia hopes to change all that.

So what is Loomia ? Loomia is a company that provides a set of web services for recommendations and personalization. Web publishers, media sites, and e-commerce retailers can use these web services to provide their users with similar/related items and personalized recommendations. The recommendation system makes use of a mix of collaborative filtering algorithms to provide item-to-item and item-to-user recommendations for media, retail goods, listings, and more. Loomia operates by collecting user preferences, calculating the products or items that the user might like based on these preferences, and send this information back to the website which can then be displayed as the recommended items to the user.

To install Loomia on your site, one has to first sign up with them at loomia.com. The site then provides you with a set of widgets or tools (javascript) which have to be deployed on the all the pages of the website. These widgets are provided by Loomia along with the deployment instructions. Once this is done, the website is hooked up to the recommender service and all the user preferences are fed to it along with the catalog of products/services offered by the site. Obviously, Loomia cannot be customized to better suit the product/service line of your online business. On the other hand, you don’t have to spend a fortune in creating a customized recommender system. Loomia is a service to which anyone can subscribe for a comparatively small fee.

Is it good enough?

Loomia.com lists betterpropaganda.com as one of its “happy customers”. betterpropaganda.com is a site that deals with music from artists who have agreed to put their work on this site and let people listen and download their tracks for absolutely no charge at all.

To learn more about you, betterpropaganda.com requests you to register and create an account with it, so it can keep track of all your searches and the songs you listen to. I did create a dummy account and decided to search for the track “Young folks” by Peter Bjork and John (Swedish indie rock band). The recommended songs were from the same genre and did appeal to my taste. The fact that this site offers very little popular music makes it very difficult to compare it to the more popular music sites like propaganda.com. But the quality of the recommendations was good enough for me, considering the fact that I had never heard of these artist or songs before visiting this site!!!

Loomia insists that it does not store any user data, which should put to rest the concerns of any privacy advocate.

Services like Loomia have now taken recommender systems from the realm of programmers and big businesses with huge R&D budgets and put them into the hands of the everyday Internet user. Loomia also shows the important role that recommender systems play in attracting customers, a role that is now being recognized as an essential component of any online service which aims to be customer centric.

Links:
1. Loomia.com
2. betterpropaganda.com

WEEK 01 - Grayboxx: Recommendations based on User Reviews

Search engine technology has evolved over the years from functioning as a tool that throws up relevant results for a given string of characters to becoming a platform that sells services and products. Search engines have always helped to point consumers to the products or services they were searching for, creating and fuelling business opportunities. But after realizing their potential as an advertising tool, they have now started to market themselves as an instrument to direct consumer interests to a business, for a small fee of course. And unlike conventional ‘in-your-face’ advertising, these companies have realized the importance of relevant and discreet advertising, making their advertising methods much more appealing and functional to the end users.

Driving all this is the fact that search engines have decided to provide a level of customization and personalization to their users not only in their search results but also in the accompanying advertising. And they are doing so in some really innovative ways, Grayboxx being a point in example of such innovation.

Now coming to Grayboxx, a search engine whose results are influenced by user reviews and are localized to a particular geographic location (based on the users selection), it looks like Grayboxx is trying to create a niche for itself in the ‘local search’ tools market. The fact that is powered by user submitted reviews takes a big burden of the users of the search engine, as they don’t have to do their own research on the how good a business is before choosing its services.

How many times have you had to look for a restaurant based on a friend’s recommendation, or search for a car dealer based on your colleague’s past experiences with the same? When you attempt to do this, you have a two step process. First, look for a relevant business that provides the service you want and then look for an online review of that business, usually on a blog or forum. What if you don’t find a good enough recommendation for that business, or even worse, don’t find a recommendation at all?

Grayboxx is trying to simplify this process by using these same user reviews as a basis for prioritizing its search results. In addition, the user can mention his/her geographic preferences, making the process of searching for a business even more relevant to the user.

CEO and founder of Grayboxx, Bob Chandra “From the start, the whole idea was premised on the fact that user reviews did a decent job in a certain number of categories when it comes to local search. But people tend not to set aside 10 minutes to write a review about a plumber or an electrician or a furniture store." Grayboxx's solution to this shortcoming lies in the patented algorithms that run it’s "preference scoring" system: "When users take actions with a business by calling or visiting, we call those implicit endorsements. What our system ultimately does is cobble these clues together, and provides a rating."

Grayboxx clearly looks like a search service geared to promote local businesses by providing favorable search results based on the past experiences of consumers with them. This raises some very important questions: “Does Grayboxx consider each and every review for a product?” … “how are these reviews graded and aggregated to get the final priority level for the corresponding business?” … “does Grayboxx give more importance to reviews by locals over those by people from outside that area?” ... and can Grayboxx avoid the criticisms and potentially deadly loopholes of recommender systems that have plagued other e-commerce ventures by companies like Walmart and TiVO.

Grayboxx has gone public, and news about it is slowly trickling in as it caters to the needs of most of the major US cities and some international locations too !!! And as with every product, only time will tell if Grayboxx can deliver on its promises and if its technology feeding off human reviews can actually do what Bob Chandra says it can.

Read more about Grayboxx @:

1. Grayboxx mixes up local search and user reviews

2. Grayboxx CEO on the underlying logic of the Grayboxx search functionality

3. Try Grayboxx !!!

4. Grayboxx founder Bob Chandra quoted in the Wired Blog