1-800-987-654 info@trouvus.com Trouvus Dashboard
Recommendations Solutions

Recommendations Solutions – How Important Are They for Web Services?

Before the advent of the Internet, people used to get their tips and recommendations about various items and sources largely from family members, friends, trusted network and experts. With the proliferation of online services with the overwhelming amount of content and items, information overload has become a critical issue. The large amount of availability of choices, while beneficial, also creates problems for the users.

The various problems a user encounters due to information overload include time wasted in search for appropriate selection, making poor choice while there are better options available and difficulties in arriving at the right choice. Recommender systems are the tools that are aimed at alleviating the information overload related issues. Today, most users largely depend on the web service itself to provide recommendations of the items that would arguably reflect users’ personal choices. The concept of recommendations solutions or recommender systems (RS) is not new. Some early websites have used them actively and over time they have become a necessary element and an integral part of any large to medium e-commerce, social networking sites and web services.

Recommendations solutions are meant to produce a list of items or products to website users based on user’s profile, online behavior, item’s profile, click-through and various other attributes. There is a fundamental difference between a search engine and a recommender system. When the user knows or at least can guess what keywords or terms needed to apply in order to find an item, a document or a product the person is looking for, she uses a search engine. In the case of a recommender system, the list of recommendations is generated for items, documents, products, etc. without an explicit search conducted by the user.

From user’s perspective, this is wonderful, because it allows the user to discover new content corresponding to her taste that she was not aware of earlier. By providing relevant but new choices, recommender systems make user interaction with the sites interesting and significantly enrich user’s experience. Web services implement recommendations solution for multiple reasons.

Firstly, an e-commerce site would like to sell more products, or services and RS helps a site to increase its conversion rate by transforming more visitors of the site to buyers.

Secondly, RS contributes to cross selling – sale of more types of products or services to the buyer. For example, an e-commerce site of electronic products may provide the buyer of a camera a precise recommendation list of lenses, memory cards, and cases. These items, possibly, would have been difficult for the buyer to discover without the help of the RS. Thirdly, data collected from the RS can help the web service to organize and optimize its content and items and improve inventory control. And most importantly, RS increases users’ satisfaction and their loyalty. RS has the capability of consistently delivering relevant, new and serendipitous items through an effectively designed interface. It can often predict the next steps of a user successfully and provide clear calls for action. Most users highly value interaction with a website with these capacities. An extended user-experience supported by the RS in turn facilitate converting the user to a loyal customer.

Recommender systems use myriad of techniques and algorithms to calculate and deliver recommendations. In its most basic form, many websites offer a non-personalized recommendation list of most popular items. The rationale behind this approach is if an item is liked by a large number of people, there is a high probability that others may also like it. Some systems perform predictive analysis and deliver recommendations by computing and comparing the user’s assigned ratings to the items she liked or disliked with the ratings of all other users of the system. The assumption here is that if the users rate same items with similar ratings, their interests are quite congruent. It is safe to recommend an item that the user did not rate but others with similar interest have rated it highly. Another method is to elicit recommendations to the user of an item based on what others have bought after searching or purchasing the same item. In more sophisticated RS extensive knowledge about the users, items, context and various other complex attributes are used.

Often, the systems are dynamic and consider the preferences of the user in question in real time. Rating of an item ascribed by a user is the most commonly used data in recommender systems. Often, web services consider that ratings are the most important indicator for a recommender system. In reality, this is not the case. The reason is people feel necessary to ascribe a rating to an item when it induces a strong positive or negative emotional link. People without any strong opinion about the item tend not to rate anything.

As, usually, small percentage of the population prefer to rate items, for many items the quantity of ratings are not representative enough. As a result, these items are difficult to cover in the computation of a recommendation list. So, although the ratings are one of the good indicators for recommendations it should not be the only one. One method of compensating this problem is to observe user’s behavior. If a customer purchased a product or watched a movie, and the system has the ability to monitor the behavior of a large number of customers, the data collected from this produces useful clues, which can become a viable attribute for the recommender algorithm.

Recommender systems today are highly sophisticated technology that applies machine learning, statistical methods, artificial intelligence agents and other state-of-the-art tools. Web services interested in implementing RS to their system should analyze, compare and evaluate the algorithms and systems and select the one suit their purpose best.

As mentioned earlier, Recommender systems increase revenue for a web service in multiple ways. They help clients of the web service to discover items that they would like to purchase. They facilitate conversion of visitors into clients, increase cross-sell by recommending complementary items and bolster development of a loyal customer base by improving the user-experience.

This Post Has One Comment

  1. anna nova

    Very informative blog article.Really looking forward to read more. Really Great.

Leave a Reply

Your email address will not be published. Required fields are marked *