 | The Business Intelligence Tools Spectrumby Mark J. Max, Managing Partner |
The business intelligence (BI) technology space is estimated to exceed $4 billion in 2003. In recent years, there has been both consolidation in the industry and new products evolving. Selecting the right BI tools and technology is a confusing process. There are a wide array of products sold in the BI technology space. Many of these products appear similar, but in fact, are fundamentally different in both capabilities and technical approach. Often we see organizations evaluating a short list of BI tools that are not suited to solve the same problem and whose underlying architectural requirements have implications that are not well understood. These differences are important to understand if you want to select the BI technology that will best serve your decision support objectives.
Understanding the BI Tools Spectrum
The first and most critical insight is the realization that there is a spectrum of business intelligence (BI) tools. Figure 1 is a diagram that illustrates the BI spectrum range.
 |
| Figure 1: BI Tools Spectrum |
Relational Reporting Tools
To the far left on the spectrum are relational reporting tools that enable reporting against relational databases. Crystal Reports is probably the best known of these types of tools. While many would argue that Crystal Reports is not a BI tool, Crystal markets this technology in the BI space, and as such, it gets consideration as part of an overall BI solution. In general, relation reporting tools simply deliver information, but provide very little if any analytical capabilities. They are inherently designed as a technology to produce reports and not as a tool to support interactive analysis.
Relational Query and Analysis Tools
In the center of the spectrum are the relational query and analysis tools. Like the relational reporting tools, relational query tools deliver data from relational databases. The difference is that they are designed primarily as interactive tools and provide some level of analytical capability. These tools are by design very open and can support all of the major relational databases. While many of these products provide varying degrees of analytical capability, they generally do not provide a separate analytical calculation engine and are inherently limited by the capabilities within SQL. While relational query tools generally do not impose data architecture, pre-aggregation of data may be required for certain applications in order to achieve acceptable query performance.
OLAP Tools
At the far right end of the spectrum is OLAP, which stands for on-line analytical processing. OLAP technology is a highly specialized technology that presents information in a multidimensional format and is designed to provide more powerful analytical capabilities than SQL. OLAP technologies use either specialized databases or specialized analytical calculation engines to support multidimensional analytics that are often difficult to achieve using SQL. While OLAP technology delivers fast performance and improved analytics, it is not a generalized reporting tool and is essentially limited to multidimensional presentation (i.e., transactional/tabular reporting and OLAP are fundamentally incompatible). Some products do provide "drill though" or "drill to detail" capabilities which enable drilling into relational tables to access more detailed data.
The following paragraphs discuss five key factors that must be understood and considered in selecting an appropriate BI technology.
1. Data vs. Information vs. Knowledge
There's a big difference between data and information. Data is defined as raw facts that have been collected, processed, stored, but not organized to convey meaning. Information is defined as a collection of data organized in a manner to be meaningful to a recipient. Business people typically want information, not data. Often, information requires transformation of data into metrics and dimensions that do not exist in the raw data model. The left end of the spectrum can support data and information; however, the burden is on the designer/implementer to transform the data into information. The right side of the spectrum is inherently focused on dimensional information, and as a result, the implementation process tends to drive the transformation of data to information.
Knowledge is defined as information combined with understanding, experience, accumulated learning, and expertise relevant to a problem, decision, or process. Knowledge is the fundamental objective of business intelligence. That is, to provide information and analytical capabilities to business people with the insights to solve problems. Whether you use relational or OLAP technology, you need to understand the transformations required to go from data to information, and consider those transformations in selecting BI technology.
2. Casual Users vs. Power Users
We often see BI selection processes where there does not appear to be consideration of the wide range of user skills and needs. In reality, all organizations have many people who need information, but generally there is a much lower number who a) actually perform analysis and b) have the technical and analytical aptitude to generate their own information and perform complex analysis. These two different audiences have fundamentally different information needs, analytical capabilities, technical aptitudes, time constraints and level of insight into the underlying business data. Based on my personal experience, 80% - 90% of information consumers are casual users who have limited time, limited expertise or limited capabilities when it comes to analysis. In fact, many BI tools have analytical capabilities that exceed the skills and capabilities of the typical casual user. As such, it is critical to match the BI tool to the needs and capabilities of the target users.
3. Reporting vs. Analytics
There is a big difference between reporting and analytics. Most likely any data warehouse implementation will need both. Some BI solutions offer very little in the way of analytics but are very capable reporting tools. The confusing market leads vendors to extend their products (in either direction) to cover a broader spectrum of the market. Before you select BI technology, you need to understand the analytical and decision support needs of your users, map those needs to analytical functionality in general, and finally be able to assess analytical capabilities in the products. Advanced analytics involves dimensional rankings, exception based selection, time series functions and other similar functionality. While analysts and power users often need these capabilities, the average casual users either may not require advanced analytical functionality, or may not have the skills to use such functionality.
4. Relational vs. OLAP Technology
There has been an ongoing debate within the BI community as to which technology approach is best: relational or OLAP technology. Relational databases are open, flexible and can serve many objectives, while OLAP technology tends to deliver better analytics and faster performance. OLAP technology requires another level of meta data and data staging, but in many cases, relational BI technology also requires additional meta data and staging of data.
To complicate things further, there are many OLAP architectures to choose from: 1) MOLAP - where data is stored in a separate OLAP cube or database; 2) ROLAP - relational OLAP where the underlying data resides in a relational database and a specialized analytical calculation engine is used to support the analytics; and 3) HOLAP - or Hybrid OLAP where some of the data stays in the relational database and some subset of aggregates are stored in an OLAP database. MOLAP will almost always perform faster than ROLAP, but requires additional processes to build cubes and may require additional disk space. Additionally, analytical capabilities are sometimes less robust in ROLAP technologies that rely on SQL instead of a separate analytical engine.
Also confusing is the distinction between multidimensional presentation and OLAP technology. All OLAP is multidimensional, but not all multidimensional presentation is OLAP. There are BI tools that provide relational based information access with no specialized calculation engine and can present information in both a tabular and cross tabular format. Generally, relational based tools with multi-dimensional presentation capabilities will not provide as robust analytical capabilities as an OLAP technology, but offer more overall flexibility and serve a wider spectrum of needs.
Each technology has its place in building decision support solutions. In reality, there are four things that you should expect:
- Any given vendor will tell you that their approach is the best;
- Most large applications will require some pre-aggregation of data to meet acceptable performance requirements;
- There are trade-offs in any decision you make; and
- The right answer for you depends on your situation and your objectives.
5. BI Product Framework and Architecture
Providing business intelligence involves two key ingredients: the front-end user interface and the back-end data repository or engine. So another key element of the puzzle is placing vendors in the appropriate categories of front-end, back-end or both components of the BI tools spectrum framework. The evolution of BI technology market forces has taken shape and the vendors can be classified into three categories:
The vendors that provide complete solutions consisting of both front end and back end components can offer one stop shopping for reporting and analysis capabilities. These single solution vendors generally use closed proprietary technology. Although some of these vendors claim to be open, in practice, customers seldom use a multi-vendor mix of these types of products.
In many ways, this product decision factor mirrors the ERP debate of "Best in Class" vs. "Integrated Suite". A critical issue in your selection process is to understand the trade-offs and benefits relative to your specific objectives.
Other Architecture Considerations
Another architecture question is choosing Web versus Client-Server deployment. Most large scale BI deployments require a web interface for the majority of users. Most of the more established BI vendors, whose original offerings were client-server, continue to offer the client-server products with the web offerings. Often the older client-server products offer more robust functionality than the web based offerings, and should be considered as an alternative for power user deployment. Several vendors have built web based products from the ground up and have no client server offerings. For large scale deployments, the reduced cost of web deployments outweighs the functionality trade-offs that may exist with a web-based front-end.
A major debate associated with the web deployment model involves using a "thin client" (HTML) versus a "rich client" (JAVA, Active-X) interface. There are clearly strengths and weaknesses to both models. The rich client interfaces tend to be easier for interactive analysis since the support "drag and drop" capability. HTML interfaces, while easy to deploy, require multiple steps to accomplish functionality such as drill down, rotation, filtering and other analytics. With HTML, the "simple functionality" is simple, but the complex analytical functionality can be challenging for casual users.
A final consideration regarding web based product offerings is "open component architecture" versus "self contained products". Some web-based BI vendors offer capabilities that can be easily imbedded in other web applications as components, while other vendors offer self contained web environments that do not lend themselves to this component approach. If you are embedding BI capabilities into a portal framework, make sure you select a technology that offers a component architecture.
Conclusions
There is no magic BI product that does it all! When selecting BI tools and technology, understand your user base, information needs and objectives before you do anything. It makes no sense to compare products that are fundamentally different and that address different needs. For a large scale enterprise deployment, both relational and OLAP technologies are often required to address all of the data-driven decision support needs. If you are targeting a single front-end product for both casual users and power users, expect to face some difficult trade-offs in simplicity versus power.
Before looking at vendors, understand which product type is best suited for you needs. Then select vendors with comparable products that fit the profile of your requirements before you begin the assessment. If you understand the BI spectrum and use it in selecting potential vendors, you will be much more likely to succeed in your deployment of a data-driven decision support system.
About the Author
If you have any comments or observations, you can contact Mark Max by e-mail at mmax@istrategyconsulting.com. Mr. Max is the Managing Partner at iStrategy Consulting, a firm specializing in business intelligence, data warehousing and analytical applications. He is a CPA with 20+ years of consulting and industry experience focused on business information and technology. He has also been a guest lecturer at the Smith Graduate School of Business at the University of Maryland. For the past 12 years, Mr. Max has designed and implemented numerous data warehouse and analytical reporting solutions for various companies and organizations. Prior to founding iStrategy Consulting in 1999, he held positions at Price Waterhouse Management Consulting, Carefirst Blue Cross Blue Shield and Oracle Corporation. Mr. Max has an MS in Business Information Systems Management and a BS in Accounting from the University of Maryland. Mark has his CPA, MCP, CISA, CCP and CSP certifications. You can get more information about Mr. Max by visiting www.istrategyconsulting.com.
Citation
Max, Mark J., "The Business Intelligence Tools Spectrum", DSSResources.COM, 05/31/2003.
Mark J. Max provided permission to archive this article and feature it at DSSResources.COM on Monday, March 3, 2003. This article was posted at DSSResources.COM on May 31, 2003.