Different recommendation scenarios provide recommendations for different web-pages. For example Scenario "product_page" provides recommendation for a page with detailed product information. When recommendations are fetched using the "product_page" scenario the recommendation engine has to know on what product detail page the recommendation should be presented. This information is called scenario context and must be provided for every recommendation call.
There are three types of the scenario context:
No context. The scenario is context independent. The best example is the global best sellers. Best sellers depend neither on the current user nor on the page the user is viewing right now.
Such scenarios can be used with other types of context (see below). The context will be ignored in this case.
Web page context. The scenario depends on the context the user is currently in. For example if the user is viewing a wrench the bolts and nuts set can be recommended. The wrench is the context item in this case.
It is possible to provide multiple items as a context. Using multiple context items allows to provide recommendations for a list of items (e.g. in a category overview) or the current shopping cart. If multiple items are provided as a context, it could be the case that not all the items are used for calculating the recommendation. Multiple items for the context are selected randomly from the provided list if 5 or more context items are submitted.
Customer history context. Recommender engine stores - anonymized - footprints of the customer. The history of the customer can be used for providing personalized recommendations. Recommendations are provided not only for the history of the current session, but also for earlier footprints (for example the items bought last month). Therefore it is required that a user can be recognized by his or her ID (which could be a cookie or login ID).
Following history list types can be used as a context:
History lists are very tightly linked to the event types (see 8. Event Types) but it is not the same.
Effectively used items for the recommendation are semi-randomly fetched from selected history type. The latest history items are preferrably used .
Read the next chapter for more information about context support and configuration.
|See the developer space to know how to provide the scenario context for the recommendation call.|
2a. Fetching Recommendations (deprecated)