Reporting, the foundation of BI, is specifically for the type of questions that are asked on a regular basis. Reports present and document answers to the most common business questions - the questions deemed so important that you want to be able to run a report and see the information on a moment's notice. Many common business problems (such as sales are down in the East, inventory is too high in the warehouse and expenses are on the rise) are readily identified in daily, weekly and monthly reports.
Reports tend to be static documents with limited flexibility. Many are still meant to be printed and viewed in hard copy. When this happens, the information is frozen in a single state forever. Reporting has evolved and has become a more interactive experience. Most Web-based online reporting now allows you to drill down and change certain filters and selections, which gives the reporting environment more flexibility.
The general lack of flexibility in reporting, combined with the never-ending barrage of questions, led to the development of analysis tools. The two most common forms of BI analysis tools - ad hoc query and online analytical processing (OLAP) - were designed to let users ask random questions of data.
Ad hoc query and OLAP tools are often used by technically savvy, business-oriented users. The tools tend to have lots of features and functions that require training and skill to understand and use. Given the scope of business today, it is easier for a serious businessperson to make an educated guess that will direct their inquiry into the data than for a purely technical person to fumble through random questions, hoping to find an answer.
Most companies use a combination of reporting and analysis tools to help run and track their businesses. Whether a flexible reporting environment, ad hoc query or OLAP analysis is used, there are actually only a few functions that are provided within common BI tools that allow you to pose your questions: sorting, filtering, drilling (down or over), pivoting, calculating and charting.
Each of these functions can be used in combination with most BI analysis tools to navigate and investigate data. You might calculate profit as sales minus expenses, sort your branches by the lowest profit, drill down on the worst performing branch to look at the sales people in that branch and resort by sales to see the worst performing sales people in that branch. That's the kind of information that will help you make a decision.
What if you don't have an inkling of a problem or can't even make an educated guess about where a business issue may be festering? Because business problems don't necessarily appear on aggregate reports, what do you do when you don't know what question to ask? That's the drawback with all of these features and functions: You need to know the question that must be asked. This is the fundamental problem with these common BI tools.
Data visualization tools use charts, lines, geometric shapes and colors to visually represent data. When all of this is combined into a single interactive dashboard, tens of thousands of rows of data can be represented on a single screen, aggregated at multiple dimensional levels. Problems can easily be isolated because unusual things in the data automatically stand out and help direct a line of questioning.
Common data visualization tools have three notable functions to represent information:
- Use of color - the connection of colors between the related charts and the data they represent.
- Isolation and chart interactions - all related charts immediately react to selections made in the others to quickly isolate business issues.
- Detail charts to expose outliers - some of the most unique chart types are used to spray tens of thousands of rows of detailed data into the small space of a single chart, so anomalies in data stand out.
A data visualization tool's job is to point out what's abnormal or unusual in records of data. These unusual records are often referred to as outliers because they lay outside the norm. Data visualization charts are excellent for surfacing the outliers in your data. Figure 1 shows the common and the uncommon return ratios for each order to a consumer electronics vendor. To clarify, if 100 units were shipped to a retailer and one was returned, there would be a one percent return ratio. When retailers order products from this vendor in groups of 100 or more, each order usually results in one or more returned items.
Figure 1: Return ratios for orders to a consumer electronics vendor. Return ratios of greater than 1.6 would be considered outliers.
By displaying the return ratio for each order on a histogram, it is easy to see the common and uncommon return ratios. All the common return ratios fall in the bell part of the curve, that is, most returns result in between zero and a little more than one percent, but a few range as high as 3.1 percent, represented in the flat plain of data to the right of the curve. Only a scattering of orders have more than a one percent return ratio, and it is easy to highlight and focus in on just these items.
If this chart is combined with others in a data visualization dashboard, the outliers can be selected to focus on the problem. The dashboard in Figure 2 adds a bar chart that aggregates returns by product category, another by manufacturing plant and another data grid with more details about the orders. Figure 2 shows the complete dashboard with no selections. Figure 3 depicts the same dashboard after the higher return ratios in the histogram have been selected using a mouse.
Be the first to comment on this post using the section below.