The Rank Engine calculates the ActivityRank for each object in your system. The Rank Engine is run as the BEA AL Rank Engine service in Windows. To run the Rank Engine in UNIX, run the following command from the command line: <PT_HOME>/rankengine/<version>/bin/rankEngineService.sh
This section discusses:
How the Rank Engine Works
Special Tuning Considerations
The Rank Engine periodically queries various repositories in the system to gather historical activity data for each object. After saving this data, the Rank Engine uses the settings that you configure on the Rank Factors page to calculate the ActivityRank of each object.
In Pathways, "value flows from one object to another". This statement means that the Rank Engine uses the ActivityRanks of source objects to calculate the ActivityRanks of target objects (where the source object performs the action on the target object). For example, because John Hill (source object) authored (action) a Sales Spreadsheet document (target object), his ActivityRank impacts the ActivityRank of the Sales Spreadsheet document.
Note that items and people can be both source objects and target objects. Using the example, the ActivityRank of the Sales Spreadsheet document (source object) impacts the ActivityRank of John Hill (target object). If nobody views, edits, or tags the Sales Spreadsheet document, it will most likely have a lower ActivityRank than documents that have been viewed, edited, or tagged a number of times. In this case, the Sales Spreadsheet document is the source object and John Hill is the target object. Note that even though the Sales Spreadsheet did not perform a true action on John, Pathways registers an action from the spreadsheet to John in order to calculate the spreadsheet's influence on John's ActivityRank.
When working in the Rank Factors page, you would tune The rank of the author in the Items area to adjust the amount of value that John Hill's ActivityRank will contribute towards calculating the ActivityRank of the Sales Spreadsheet document. You would tune The rank and number of pages authored in the People area to adjust the amount of value that the Sales Spreadsheet's ActivityRank will contribute towards calculating John Hill's ActivityRank.
Take careful note of these points when tuning the Rank Engine:
When you tune any rank factor, you are adjusting
the influence that the ActivityRanks of the source objects have when calculating
the ActivityRanks of all relevant target
objects.
For example, adjusting The rank of
the author adjusts the ranks of all items that have been authored.
In other words, the ActivityRanks of the authors impact the ActivityRanks
of all items that they have authored.
The tuning of rank factors is relative; this is the case no matter how many rank factors affect that item or person. When you tune a rank factor, you are adjusting its influence relative to the other rank factors that impact an item or person. The following table lists the rank factors that impact the ActivityRanks of Item A and Item B (tunings are in parenthesis):
|
Item A Rank Factors |
Item B Rank Factors |
|
The rank of the author (75%) |
The rank of the author (75%) |
|
The rank and number of people who edit (85%) |
The rank and number of people who edit (85%) |
|
The rank and number of people who tag (95%) |
The rank and number of people who view (15%) |
As you can see, the ActivityRank of Item A is most influenced by The rank and number of people who tag, which is set to 95%. The ActivityRank of Item B is most influenced by The rank and number of people who edit, which is set to 85%. Although The rank and number of people who edit is tuned to the same position globally, it has more influence on the ActivityRank of Item B than it has on the ActivityRank of Item A.
In order for your rank factor tunings to take effect, you must use the Schedule And Status page to recalculate ActivityRanks. In many cases the Start Over and skip the data gathering phase option is the most efficient way to update your system's ActivityRanks.
Try to adjust as few rank factors at a time and adjust those rank factors in small increments. Depending on the history of actions in your system, making adjustments to the rank factors might greatly impact the ordering of result sets.