Building High-Performance Business Intelligence (BI) Systems

Business Intelligence.

Companies that understand the importance of business intelligence design an information architecture that includes data warehousing, solid transactional data, and powerful analysis tools.

 

Finding the right formula for turning data into bottom-line results has been more elusive for some companies than for others. A study by the Andersen Consulting Institute for Strategic Change describes how the refrigerated dough division of Earthgrains increased earnings 70 percent by using SAP data to analyze product profitability. However, other companies have poured millions into business intelligence (BI) initiatives that have failed to produce significant results. For BI to succeed, companies need a full suite of analytics tools, which surround one or more data repositories.

 

Information Storage

Historically, most companies have approached business intelligence with the goal of gathering all important enterprise data in one central data warehouse that users throughout the organization can access. Increasingly, though, companies are running business intelligence systems off of smaller information repositories, datamarts that focus on a specific business process or function.

 

These targeted databases let businesses take a more incremental approach to building a data warehouse. They also reduce the time extracting information takes. One datamart might focus on sales, another on operations and another on financials. The data marts are easier to manage, they can be built more quickly than the galactic [data warehouse], and they allow for a gradual rollout of data warehousing capability.

 

Companies must seriously evaluate how much data they need to store in each repository. Even within one area of the business, not all information is equally important. One way to decide which data goes in the datamarts or data warehouse is to create a ranking system. For example, a data warehouse that will store information to support customer relationship management can rank its different types of data based on their usefulness in predicting customer behavior.

 

Whatever data a company stores in its information repositories, it must make sure that it uses only high-quality transaction data. If your transactional data isn’t current, consistent, accurate and complete, then why to build a data warehouse to put on top of that poor information?. All you’ll end up doing is analyzing bad data.

Data Extraction

A second key element of a BI system is the software that enables users to pull out of the data repositories the bits and pieces of information they’re interested in. Data extraction tools are separate from the data warehouse or datamart, and they must be evaluated separately. Businesses new to BI commonly make the mistake of thinking that one extraction tool fits all. Companies are looking at competitors and figuring, ‘Well, they’re using tool X, so that’s what we’ll use,’ but different data warehouse extraction tools are designed for different purposes.

 

There are two basic types of data extraction tools. Query software answers user questions by accessing information in a large, relational database management system (RDBMS), such as a data warehouse. Alternatively, online analytical processing (OLAP) tool can extract data from a smaller, multidimensional database (MDDB) — also called an OLAP cube — that has already calculated responses to all potential user queries. OLAP tools are patterned after the concept of a spreadsheet so that the user can view data in different dimensions — for example, profitability by product, by geographic region or by the office.

 

A query tool is usually the best solution for ad hoc reporting and analysis, while an OLAP tool is preferable for modeling and forecasting. A query tool on RDBMS is not set up to store information or do what-if scenarios, and it can take up to two or three minutes to respond to a question from a user. On the other hand, the OLAP tool on MDDB reacts immediately to a user question, and it can do what-if scenarios. However, RDBMS databases are more flexible. For users who don’t know what they want to look at, they can use the query tools to make ad hoc queries and do discovery analysis. Once they find what they want, they can generate a report that runs that information and then put it into a datamart for regular reference.

 

Although an MDDB system is faster once it’s ready for queries, it is slower to get up and running than an RDBMS database because of its up-front calculation process, which often takes several hours. Also, MDDB databases face constraints on how much information they can store, whereas the size of the data warehouse only limits RDBMS query applications.

 

The size of the data repository a user needs to query may have a bearing on the type of extraction tool that would work best for that user. For a large data warehouse, it often makes more sense to use the [RDBMS] tool because you don’t have the storage issue, whereas with a datamart the multidimensional tool may be more appropriate.

 

Many businesses have only one type of data extraction tool, but to maximize employees’ ability to leverage the information in a corporate data warehouse, companies need both the query tools that run on an RDBMS database and multidimensional OLAP tools. There are many ways to get the data out of the data warehouse, but finding the right approach depends on what the user wants to do with the information. For instance, if you want to report on the consolidated financials of your company, look for a data warehouse product that allows you to do the multidimensional analysis.

 

The Human Element

Another essential ingredient in a business intelligence initiative is having people with the right analytical skills. Data warehousing won’t do you much good unless you know what problems you want to solve and possess the human analytical skills to solve them; many companies have ignored the human factor. High-performance companies realize having analytical capability is really about human performance, not a technology issue.

 

Cisco Systems Inc. has created a work environment in which employees are encouraged to use corporate data when making business decisions. One aspect of their jobs that they must use the information to manage is travel expenses. Travel expenses are the second most controllable costs at Cisco, right behind payroll. They developed a datamart for travel expenses that help managers, for instance, understand why an employee may not have taken the lowest airfare. Employees have the tools, via the travel expense data warehouse, to access this information on the Web so that anyone can see it at any moment.

 

The Value Proposition

The cost of building a BI system is falling. Not long ago, few prepackaged data warehouse products were available. Most companies required custom solutions, which were costly. However, now a variety of off-the-shelf packages is available, which has made buying a data warehouse preferable to building one for many organizations.

 

Instead of building it from scratch, try to find a pre-built data warehouse that fits your needs. Your costs will be lower, and you’ll reduce the risks historically associated with the building. That’s why approximately 80 percent of all data warehouse projects are pre-built. If you’re a financial manager, buy an analytical application that can instruct you on its usage. It’s a lot cheaper than building your own and paying a consultant for advice.

 

What can companies expect in return for their investment? One approach is to try to measure the cost of providing answers to decision support questions in a data-warehouse-less environment by looking at the number of custom operational reports that would have to be developed, plus any additional manual processes that would be needed to combine and reformat these reports to answer a specific question. The real benefit of the data warehouse starts to become apparent after going through this a few times.

 

Building an Analytic Capability

A study by the Andersen Consulting Institute for Strategic Change called “Data to Knowledge to Results: Building an Analytic Capability” examined 20 companies that have achieved bottom-line results through business intelligence (BI). Here’s a summary of the study’s recommendations for a successful BI project:

 

Develop a data-based culture. In the study,  most successful businesses had the most pervasive data-based cultures, which they built by obtaining high-level support for the project, understanding the current culture, spurring cultural change in small steps and maintaining patience. Most companies in the study had been working on these changes for between 2.5 and ten years.

 

Develop critical skills and experience. Educate business-oriented people about data and data-oriented people about business. The study mentions that Bank of America Corp. successfully educated its marketing staff about data modeling through educational forums.

 

Carefully structure analytical resources. Are the organization’s structure and culture heavily oriented toward centralization or decentralization? Companies such as Hewlett-Packard Co. and Kraft Foods Inc. have a history of highly autonomous business units, so they would probably find it difficult to centralize their analytical resources, the study says.

 

Improve the technical environment. Let strategy drive decisions about whether to custom-build applications or purchase the prepackaged software.

 

The study concludes by emphasizing that companies which succeed in data-derived analysis make it part of the organization, align it with their strategies, gain senior executive support and focus on solving specific business problems.

 

 

Michele Y. Thompson is an author, contributing writer on MyStock911 and MortgageExpertGuide , commercial mortgage broker,  entrepreneur, and finance coach.  The culmination of her work in mortgage banker finance, global investor services, real estate, and debt consulting; along with her advanced degrees has driven her to help new and existing businesses reach their goals for over 20 years.