Data Warehouses and Data Marts
Central to traditional Business Intelligence efforts is a data mart or a data warehouse. Theyre both basically the same thing: a specialized kind of database thats designed to support business analytics and to contain data from one or many different sources. The difference between a mart and a warehouse lies in their scope: A data mart only seeks to serve the needs of a portion of the company, such as the marketing department. A data warehouse seeks to serve the entire company.
Creating these is one of the heavy costs associated with old style B.I. For the average SME using a current day B.I system, they are just not necessary, though Dynamic AI will work with these if already in place. Typical SME Data volumes and the power and agility of current B.I package solutions means that production databases can be interconnected and analysed with no impact on routine daily operation.
Data mining is simply the process of extracting patterns from data. Do you need to find out why the distribution center processes orders more slowly at certain times of the month? "Mine" your data looking for patterns perhaps youll find that things slow down as the supply of cardboard boxes dwindles because the box vendor isnt fulfilling their orders quickly enough.
Its important to realize, however, that data mining cant reveal patterns in data that arent present in the data being mined. That sounds obvious, but it can be deceptive because data mining can often seem to reveal patterns that arent really there. What you can find yourself looking at is a pattern comprised of symptoms rather than causes, when the causal data isnt present in the data warehouse. This isnt to say that data mining isnt usefulits at the heart of B.I, in fact. Rather, you just need to be aware of whats in the data, and what isnt, and take common‐sense approaches to verifying and validating the conclusions to which your data mining leads you.
Reporting and Dashboards
With all that data available, most users will need a simplified means of looking at commonly examined information. Reports are one obvious product of a B.I application, and they can range from high level summaries to extremely detailed analyses.
In fact, given the freeform layout of a good web based B.I system, you should be able to get from high level summary right through to detailed analysis via easy point and click drill down facilities. This is a major distinction from the paper based layout of the purely reporting packages that everyone has experience of.
Another option is a dashboard, which provides a summary of common metrics possibly from multiple sources, and often contains planning and actual comparisons, usually realised in a visually attractive, simplified user interface (UI)such as this example.
Management at a glance
Dashboards dont often drive direct decisions; rather, they let individual users get a feel for general performance, such as sales, inventory turnover, customer complaints, and so forth. Dashboards may also include trendssuch as day by day sales figures compared with goals or plans. Again, the dashboard isnt going to let you know why sales are where they are, or maintenance & repair tasks lagging behindbut if theyre significantly off target, youll be able to tell at a glance and initiate further investigation.
A well designed dashboard will let you know if everything is all right or whether further investigation is warranted and is a very useful aid to management by exception. This can be taken further by defining email or text message alerts if really critical values go out of range.
Drill up, Drill Down
Analyzing information can happen in many ways. You probably already use some form of reporting system, but more intuitively with a B.I system, it is easy to provide drill-down or drill-up within a report.
Drill-down is self explanatory click on a data value and you should get an expanded display of the data records that add up to that value. Where data is summarized into multiple levels Product Category, Product, Sales/Year, Sales/Month, Customer, Purchases, then the drill-down can explore through all these levels.
Drill-up is not quite so obvious. Drill-up is more akin to a filter and should apply the value clicked on to everything in view the whole dashboard of reports and graphs. Click on a value such as. Product in a list of Products and only details for that Product will show. Now you could drill-down into those details without being distracted by all the other product details. Next, switch the drill-up to Customer and you can explore in depth just that one customers data.