Suggestive consumer experience with recommender technology..
Guest Post by
Ketan Chinchalkar, Senior Project Manager, MFG-ADT Online, Infosys
Consumers face a dizzying array of choices when navigating through the online content and products. Traditional navigation and search can leave consumers wanting - and leave money on the table as far as e-retailers and content owners are concerned and hence it is very important that a website provides a suggestive experience to the consumer based on his behavior on the site and also through implicit and explicit preferences. This will help the website achieve customer stickiness and loyalty, better content monetization and a potential sales conversion. A "Recommendation Engine" in layman's terms is defined as a platform which presents the right content, at the right time, in the right context, at the right value, over the right channel, to the right customer. Now this content can be an e-commerce item, a product, part or a service, offers, promotions, pdf and office documents, a marketing campaign, news, promotions, images, videos, audio, social comments, ratings, reviews, community content and blogs content etc. It will do wonders, if similar kind of suggestive experience is provided in a multi-channel environment like email, mobile, print, POS, stores, kiosks, customer service, social networking and that too by mining and leveraging cross channel user behavior.
A typical recommender system will take inputs in form of User (profile attributes, demographics, clickstream, transactions, and searches), Content (products, services, offers, promotions, metadata, and media) and the User Context. The core of recommendation engine is real time data and pattern mining algorithms, Artificial Intelligence, Neural networks and most importantly the collaborative filtering techniques. The data mining Engine creates and updates relationships between like-content based on the content taxonomy or in some engines it may be based on customer profiles as well. Data mining algorithms analyze customers' browsing and shopping history, answers to user-preference questionnaires and surveys, and profiles resulting from this information, and translate consumer activity into product and preference relationships that could indicate what consumers might be interested in. Collaborative filtering involves grouping of users having similar preferences and browsing history and then create multiple associations of item to item and item to user recommendations. For e.g. similar customers who viewed this item also viewed, Users who bought this item also bought. Concept wise, the output of a recommendation engine is classified into 3 types: personalized based on user profile and browsing behavior, social based on similar users profile and transactions and related based on the relation and similarity of the content. Practically, no one approach will meet all the needs, devoid of shortcomings and hence the recommender systems are designed to mix the standard approach appropriately to enhance the relevance and the context quotient. Recommender systems also have a business rules or a filter engine for the business users to control the recommendation output and filter the recommendation output in multiple dimensions like user, items and transactions.
In the recent years, Social media has been enjoying a great deal of success, with millions of users visiting sites for social networking, blogging, micro-blogging, sharing etc. These social media sites rely principally on their users to create and contribute content; to annotate others' content with tags, ratings, and comments and to form online relationships. Facebook is playing a big role in today's social word of mouth and is indeed becoming a trusty recommendation engine. Facebook users are not only creating a more personal relationship with a brand, they're sharing that relationship with their friends and family. Many website catalogs/shopping carts can now be integrated with the social networking sites, like Facebook etc. and the user interactions there can be a valuable source of input to these recommender systems. The collaborative factor gets implicitly taken care by the social grouping in these sites and if the consumer likes, dislikes, tags, comments , opinions, bookmarks, blogs, posts, shares, polls etc. are being inputted to the recommender systems in an offline manner it can provide a real personalized and social perspective to the recommendation output. What I mean by real is, the recommendations derived using the social media source are more of value to the user, as the similar users based on which the recommendations are provided are his close friends, colleagues, acquaintances and relatives. The Social Web provides huge opportunities for recommender technology and in turn recommender technologies can play a part in fuelling the success of the Social Web phenomenon.
In the overall technology landscape, a Recommendation engine can integrate with the Search Engine where the search engine acts as an aggregator of the User and Item content across various enterprise repositories. It can also interface with the Search Engine for the search keywords and the search term based recommendations. Recent trends have shown that cases where real time recommendations are not needed the user clickstream needed as one input to the recommendation engine can come from the Web Analytics feed. Business users also take the advantage of the Campaign Management/Lead management systems, Data warehouse and CRM systems to create targeted groups or control groups based on the segmentation done within these systems. Recommender output can also be integrated with email, SMS and print Campaigns which have an amount of personalization and cross sell/up sell.
The recommender technology is still not completely mature; organizations are not 100% sure if implementing recommendations would increase the sales revenue/conversions marginally, as the ROI can be difficult to measure. Lot of Recommendation Engine products in the market are new and emerging in this technology and will be mature as more implementations happen across the globe. There are also many open source recommendation algorithms available in the market which can be leveraged to custom built a full blown recommendation engine. Amazon, the most popular B2C and B2B e-talier has its own patented Recommendation platform and uses traditional collaborative filtering, cluster models and search based filters to personalize the online store for each customer. Most of the commercial Recommendations systems in the market are available as hosted service at the vendors end, and very few provide an option of co-locatable product at the customer side. As there are no clear cut market winners in this technology, which tool or platform to go with OR whether to build the engine from open source algorithms, will depend on the customer specific requirements/use cases, pricing model, TCO and time to market. Most vendors provide simple APIs so that their engines can easily integrate with websites, emails, stores, mobile devices etc. Some vendors offer remote business rules management, reporting, and A/B testing capabilities as part of a packaged recommendation service.
To say a recommender technology only is applicable to a retail domain will not be true, as now days many manufacturers/OEM have their own e-commerce/content/brand sites and they are also now becoming internet and social savvy. For example a recommender system in below use cases will also give a personalized experience to the consumer.
· Content/Brand/service provider website for providing real time personalized targeting
· e-commerce website for providing personalized, social and related item recommendations
· Customer service, Call center solutions to give a personalized experience to the customer during a call, chat or an enquiry.
· Personalized experience in Knowledge Management solutions
· Campaigns (Web, Email, Mobile, Print)
· In store devices, Kiosks
· Improving Product configurators by means of a collaborative recommender system
· Social commerce
· Many more use cases where there is user-item association
In the near future, I see a recommender technology becoming a ubiquitous piece of multi-channel consumer experience. For example, the technology could leverage a location based component. "I might have a wireless communications device with GPS capability", and as I get near a store, I could get pinged with a message saying there's a sale there on a CD I might like." One burgeoning development is matching consumer tastes across different lines of businesses, such as using knowledge of customers' tastes in one area, like music, to sell them products in another area, like books. This will also depend on future business alliances and partnerships, along with advances in the technology.