An analytic application lets you facilitate an entire business process, from analysis to decision making. Such a process requires collaboration between the people who have the domain expertise and the people who make the decisions, and often many people need to share their work and conclusions. Because collaboration is an integral part of a business process, any complete analysis application should include a way to facilitate this sharing of ideas and work. SQL Server 2000 Analysis Services offers simple collaboration features in the form of what-if analysis—for example, you can use Analysis Services to perform cell-level and dimension-level write-back. But effective collaboration requires more than simple write-back capabilities; your application should support the ability to share business logic, conversation threads, and action plans. Let's examine a proposed architecture for a collaborative analytic application that uses Analysis Services and look at some source code you'll need to get started.
Sharing Business Logic
Because Analysis Services doesn't have built-in capabilities for sharing business logic, conversations, and action plans, you need to take advantage of other Analysis Services functionality to build these features. First, let's look at how to share business logic. In the following example, I use the term business logic to mean business-specific formulas such as calculated members or named sets. An example of business logic is the formula you use for determining your top 10 customers; such a formula is most likely business-specific. In some businesses, the most important criterion for determining who is a top customer might be the revenue the customer generates—in other kinds of businesses, the most important criterion might be profit or longevity.
Analysis Services can store and retrieve calculated members or named sets, but it has no facility for creating new server-based calculated members or named sets through the client-programming interface ADO MD. To create new, shared business logic on the server, you must use Analysis Manager or the administrative programming interface Decision Support Objects (DSO). (For information about Analysis Services' three programmatic interfaces, see the sidebar "Analysis Services Interfaces," page 42.) Although DSO access requires OLAP administrator privileges, you can develop a service to deploy on your application's OLAP server (or middle tier) that has OLAP administrator access. You can then make this service available to client applications so that the applications can perform actions such as creating new business logic.
Document Sharing
The next collaborative feature you might want to implement is document sharing. The documents you share can include conversation threads and action plans associated with information that your analysts discover in the cube data. For example, when an analyst is viewing profit for the last few quarters, she might discover that a particular product line is losing money. The analyst then develops an action plan associated with improving profit for that product line. After sharing the plan with others who view the same information, the analyst might also want to start a conversation thread that explores other possible improvements.
Because the analyst's action plan and conversation are associated with a certain portion of the OLAP cube, storing the information with that portion of the cube makes sense. This lets other analysts see the information only in context. You can link information to a portion of a cube through an Analysis Services feature called an action. An action is a link to information or a command that's associated with cube metadata such as a dimension, a level, or a cell. Actions are similar to calculated members and named sets in that you can't create or edit them through the client API ADO MD unless they're available only to the client that created them. In other words, you can't create shared actions. In my proposed architecture, I want to create and edit actions for use by more than one cube user, so I need to use the administrative API DSO.