In Chapter 1, we saw how business intelligence is used to support effective decision making. It provides foundational information on which to base a decision. Business intelligence also provides us with feedback information that can be used to evaluate a decision. It can provide that foundational and feedback information in a number of different ways.
In some cases, we know what information we are looking for. We have a set of particular questions we want answered. What is the dollar amount of the sales or services our organization is providing in each region? Who are our top salespeople? In some of these situations, we not only know what we are looking for, but we also have a good idea where to find the information when we design the business intelligence solution.
When we know the question we want answered and have a good idea where that answer is going to be found, we can use printed reports to deliver our business intelligence. This is the most common form of business intelligence and one we are all familiar with. For many situations, this format works well.
For example, if we want to know the dollar amount of the sales or services provided in each region, we know where to find this information. We can design a report to retrieve the information and the report will consistently deliver what we need. The report serves as an effective business intelligence tool.
This is an example of layout-led discovery. With layout-led discovery, we can only learn information that the report designer thought to put in the report layout when it was first designed. If the information wasn't included at design time, we have no way to access it at the time the report is read.
Suppose our report shows the dollar amount for a given region to be unusually low. If the report designer did not include the supporting detail for that region, we have no way to drill into the region and determine the cause of the anomaly. Perhaps a top salesperson moved to another region. Maybe we have lost a key client. The report won't give us that information. We quickly come to a dead end.
In some cases, we know the question, but we don't know exactly where to look for our answer. This often occurs when the information we initially receive changes the question slightly. As in the example from the previous section, an anomaly in the information may cause us to want to look at the data in a slightly different way. The unusually low dollar amount for sales or services provided in a specific region led us to want detailed numbers within that region.
In other cases, we know where to look, but it is not practical to search through all of the detailed information. Instead, we want to start at an upper level, find a number that looks interesting, and then drill to more detail. We want to follow the data that catches our attention to see where it leads.
This is data-led discovery: the information we find determines where we want to go next. The developer of this type of solution cannot know everywhere the report user may want to go. Instead, the developer must provide an interactive environment that enables the user to navigate at will.
To implement data-led discovery, we need some type of drilldown mechanism. When we see something that looks interesting, we need to be able to click on that item and access the next level of detail. This is, of course, not going to happen on a sheet of paper. Data-led discovery must be done online.
In some cases, our data may hold answers to questions we have not even thought to ask. The data may contain trends, correlations, and dependencies at a detail level, which would be impossible for a human being to notice using either layout-led or data-led discovery. These relationships can be discovered by the computer using data mining techniques.
Data mining uses a complex mathematical algorithm to sift through detail data to identify patterns, correlations, and clustering within the data.
Where layout-led discovery and data-led discovery usually start with summarized data, data mining works at the lowest level of detail. Highly sophisticated mathematical algorithms are applied to the data to find correlations between characteristics and events. Data mining can uncover such nuggets as the fact that a customer who purchased a certain product is more likely to buy a different product from your organization (we hope a product with a high profit margin). Or, a client receiving a particular service is also likely to need another service from your organization in the next three months.
This type of information can be extremely helpful when planning marketing campaigns, setting up cross-product promotions, or doing capacity planning for the future. It can also aid in determining where additional resources and effort would produce the most effective result.