Wednesday, February 5, 2014

Make the right decision using a data warehouse

In my previous blog, I discussed about the goals and benefits of a DW/BI system, the primary goal being aiding an organization in converting knowledge into profit through improved decision making. According to Ralph Kimball, ‘the real purpose of a data warehouse is to be the perfect platform for decision-making’. So, how do you make decisions using a data warehouse? How do you solve operational problems faced by organizations?

First, you need to compartmentalize the problem into segments each of which can be conveniently dealt with. Next, you need to identify the principal activities of the organization for which you are trying to solve the problem. Asking the business user questions like ‘What is it like to be a manager of a large company?’ and ‘What is it that you do when you do your job well?’ help in compiling a list of the principal activities and uncovering the Key Performance Indicators (KPIs) used to evaluate the success of an activity. Some examples of KPIs are revenue (marketing), equipment utilization (manufacturing), brand awareness (advertising), etc. KPIs are different wherever you go but it is important to identify them by developing the skill of listening even when you don’t know the user’s job well. These KPIs will give us a good start to designing the data warehouse.

The classic steps of bringing information and making it useful to the users for decision-making (as proposed by Bill Schmarzo) are:
  • Publish data, publish reports
  • Identify exceptions in the current business operations
  • Determine the root causes for the identified exceptions
  • Provide decision alternatives
  • Track actions which follow the decision

Step-3 is the secret of data warehouse/decision support system: ‘what do the users want to do when they ask the question - Why?’ Suppose you tell the marketing manager (user) of a large retail firm, ‘the revenue (which is a KPI) for this month is low in North America’, he will answer ‘drill down and show me the specific regions where this exception has occurred’. Drilling down is the most fundamental response to step-3. It will reveal the specific regions where the revenue is low.

Another possible answer by the user will be: ‘drill across and show me the various factors that could have influenced this exception’. The possible factors include performance of the organization's competitors in that month, performance of the sales team, etc. Drilling across provides data from other data marts in the organization existing at the same point in time. 

While designing a data warehouse, you need to anticipate drilling down and drilling across multiple data sources beforehand. This is where conforming dimensions in your data warehouse will play a huge role - you will have an integrated view of your business and the ability to analyze data from different areas of your business. It will also help you in understanding what to do when a KPI is abnormal. All in all, you will be well-equipped to make the right decision!

Thanks for reading. We will explore another interesting DW/BI topic next time.


References:
1. The Data Warehouse Toolkit, 3rd Edition - by Ralph Kimball  & Margy Ross
2. http://www.kimballgroup.com
3. http://en.wikipedia.org/wiki/Key_performance_indicator

No comments:

Post a Comment