ChoiceStream RealRelevance is a hosted service that provides automated, personalized merchandising and marketing based on each consumer's preferences and shopping intentions.
ChoiceStream RealRelevance analyzes the content to be recommended (e.g., general merchandise, movies, electronics) as well as the context in which recommendations will be delivered to provide the most accurate, useful and relevant recommendations possible.
Getting Started
ChoiceStream RealRelevance observes your consumers over a short period of time in order to detect patterns of behavior. The aggregated activities (e.g., purchases, product page views, clicks, ratings) performed by your consumers inform ChoiceStream about what items tend to be viewed together within a session, which products are purchased together or as add-ons, and how users interact with those items.
Once ChoiceStream understands your inventory and how users interact with it, it can provide accurate, relevant recommendations.
The Right Algorithm and Data for the Job
Collaborative filtering techniques used by most recommendation engines work well in some scenarios, but can fail miserably in many common situations. These include scenarios such as when a new item is added to inventory that has no activity related to it, or when a new visitor comes to your site about whom little, or nothing, is known. Solutions that rely too heavily on collaborative filtering will produce inaccurate, and often ludicrous, recommendations in these situations.
Rather than limiting itself to collaborative filtering, ChoiceStream automatically interprets each situation and employs the right algorithm for the job at hand. Below are just a few examples of how ChoiceStream uses a mix of best-fit algorithms as well as your customers' shopping data to successfully deliver appropriate recommendations where other services fail.
- New Users. First time visitors are automatically detected and receive 'default' recommendations, which are typically popularity-based as measured by activity data; once a user performs even one activity, ChoiceStream can personalize that user's experience.
- 'Flash' Content or New Catalog Items. ChoiceStream leverages the taxonomy or other attributes of the item as well as shopping data to generate recommendations.
- Top Selling Items. ChoiceStream can automatically detect and suppress recommendations of items you are already promoting elsewhere, or that users will likely buy anyway.
What Data Does it Need?
ChoiceStream typically collects three kinds of data to power recommendations: item data, click data, and transaction data.
The item data should include an identifier for the product or item being recommended. It should also include the fields needed for display such as title, price, image link and short description. This data is typically collected through the insertion of simple JavaScript into a Web page — similar to a web analytics integration — although it can be augmented with feeds if desired.
Click and transaction data are collected at the individual user level (although no personally identifiable information about the user is needed, nor gathered by, ChoiceStream). These data are typically collected through the insertion of simple JavaScript onto specific pages. If your site or store supports offline (e.g., brick and mortar) activity, you can augment with feeds if desired.
Once the ChoiceStream engine understands the basic characteristics of your catalog, how users interact with that catalog, and the individual transaction histories of each consumer, it can then leverage that data to deliver targeted recommendations based on each consumer's interests.