Return to Newsletter Contents...

It's All Semantics

by: Joe Roberts, Principal Consultant, BECP, SLE, MCP

The biggest hurdle to delivering effective “self-service” BI and/or Operational Analysis is the design of the semantic layer. The biggest hurdle in the design of the semantic layer is in understanding what the semantic layer must provide to the BI platform.

The semantic layer provides a foundation for answering business questions. So before we can design a semantic layer we must ask the questions. Each question is the end result of a “user story”. And as we all know, every good story answers the questions: who, what, when, where and why? Who needs to know this information? What do they need to know? When will they need to know it? Where will they need to see it? Why do they need this information?

As BI specialists it’s our job to translate business questions into data entities. To do that, we need to think in terms of dimensions, measures and filters. Each question needs to contain one or more dimensions, one or more measures and possibly one or more filters. We need to guide the business users in the process of defining and refining the questions to which they need answers. Who, when and where is easy! What and why is a bit harder. 

Each question needs a subject (measure) and a context (dimension/filter), this is our what. I often see questions coming back from business users like “What is the number of calls coming into our call center?” That’s great, that gives me subject “number of calls” but in what context? I would take the above question back to the business user and ask “Number of calls based on what?” What is the context of this question? A more useful question might be “What is the number of calls coming into our call center by day for specified products and demographic type?” In this question we have both measures (calls) and dimension (products) and possibly a filter (demographic type).

So why do we need to know the motivation? Motivation helps us understand the scope of our semantic layers and anticipate future needs. For instance:

Question: What is the number of calls coming in to our call center by help desk employee per hour?

Motivation: We need to evaluate our call volume, over a long period of time, to help us determine if we have enough trunks and enough help desk employees.

Is this an enterprise-wide initiative? Will this information be used by other groups or users? Are there additional dimensions and measures we can provide that will help in this area? Should we create a unique semantic item (Universe/Business View) or incorporate this into an existing one?

Go to Top  |  Return to Newsletter Contents