Liberate your operational data

Book Cover

The following article is excerpted from E-business Intelligence: Turning Information into Knowledge into Profit, by Bernard Liautaud with Mark Hammond. Used with the permission of the author and McGraw Hill.

Enterprise business intelligence systems liberate the operational data in enterprise resource planning applications and make it available across the organization for analysis. A company that deploys a costly and complex enterprise resource planning system usually wants to leverage intelligence from the ERP applications as quickly as possible to go beyond process automation, maximize return on investment, and extend the visibility and reach of the ERP applications.

There are generally three phases toward a full-fledged exploitation of ERP applications for intelligence:

  1. Limited access against the operational data store. This kind of implementation lets selected users access the data directly in the production ERP application. Shortly after an ERP implementation is up and running, some instant gratification is delivered to users with an urgent need to access and analyze the data directly against the data store. This solution is attractive because it delivers results quickly and does not require an intermediary data structure. It is often seen as a solution that cannot scale, because the security and performance of the ERP system is mission critical to the customer; the IT staff does not allow many end users launching long ad hoc queries directly into the system, potentially risking the functioning of the system. It is hard to get good performance for analysis activities without crippling the ERP application. In addition, the ERP system is designed for transactions, so it is very hard to build a system simple enough to use by business users. However, as we will see in the sidebar on this page, some companies have obtained huge benefits from this approach. It has proven to be a fast and efficient solution.
  2. Scheduled reports. Reports are run during off-peak hours directly against the ERP system. This solution solves the performance issue. It is limited to giving predefined, prescheduled reports to the user community and, as such, is a rigid implementation where business users cannot ask any question any time. However, it is perfect for a large group of users who want to receive predefined reports and do not want to perform their own analysis.
  3. Design and deployment of a data warehouse. Here, data is extracted and moved from the ERP application into a data warehouse for analysis and report generation. This is the preferred way to implement a robust system on top of an ERP. It enables the company to design the warehouse for information access by nontechnical users. The process of transformation and extraction from the ERP to the data warehouse can be managed and scheduled regularly, allowing better control of the data. Finally, it enables the creation of data warehouses that mix data coming from different environments.
Improving communication
Develop a common language, leading to goal alignment. Implementing enterprise business intelligence across a corporation's activities requires departments to agree on basic business terminology. For instance, different departments may not define a customer in the same way. Discussing and defining common vocabulary can help divisions align not just their business terminology but also their work processes.

The first part of the process is to ensure common semantics. A good business intelligence system allows you to define and store business terminology. For instance, a term like sales revenue can be complex to define. It involves pricing information, discounts, potential rebates, and the quantity of items sold—all at the same time. Members of a marketing group, for instance, should share the same definition for sales revenue. However, the accounting department may use the same term when referring to a different definition, as it distinguishes bookings from recognized revenue. The time of sale and criteria for revenue recognition are key to the definition. It is vital that business semantics be defined for different business areas. As part of this process, there must be an agreement on corporate terms that spans departments. The term customer, for instance, may need to be used consistently throughout the enterprise.

Having an agreement on business terminology goes a long way in improving the company's agility. Users from totally different departments can communicate in a much better way and share common goals. The second part of the process concerns data consistency. High-quality data and consistency among disparate data stores help ensure that decisions are made based on accurate information.

Walter Nelson, senior vice president of core engineering at Ventro, a business-to-business online trading exchange in Mountain View, Calif., is a big believer in data quality. Data quality was a large issue when he served as an IT manager at Fair, Isaac, a large credit scoring and financial consulting company in Marin County, California. "If you do not have a quality data set that is well behaved, you lack that fundamental base for analysis," says Nelson. "At Fair, Isaac, the nature of the data we received was highly variable. The data we got contained a lot of junk, and we spent a lot of money just to build the tools to scrub the data. And now at Ventro we spend a prodigious amount of energy in cleaning up our data and making it appropriate for our end users. But once we have it scrubbed and in our repository, it is an extraordinarily precious resource."

Promote accountability and efficiency. A common issue is that some departments take too long to communicate information. For instance, the finance department, a perennial whipping boy, is often faulted for not providing reports as fast as business managers like. Since enterprise business intelligence dramatically speeds querying and reporting time, internal requests can be satisfied much faster, thereby improving relationships, employee accountability and efficiency.

Stimulate curiosity. Using enterprise business intelligence across multiple steps of the value chain generates the greatest benefits. The discoveries made are often the result of the curiosity of one individual making inquiries on the borderline of his or her official job description, affirming that positive results can be obtained from giving free rein to users' autonomy and curiosity.

Successful strategies for implementing an enterprise business intelligence system
As for most strategic projects, the success of an enterprise business intelligence initiative is not guaranteed. Its success depends on a number of essential factors: its role as part of corporate strategy, the involvement of business managers in its purchase, a careful implementation, and a federated approach.

Enterprise business intelligence must be a part of business strategy
While enterprise business intelligence allows the identification of hidden costs or of new revenue opportunities, these benefits are achieved only once action is taken, i.e., costs are cut or additional revenue is pursued. The enterprise business intelligence system is part of a process, and its returns are included in those of the whole. The enterprise business intelligence project brings value as part of a larger business strategy, and the value of the project can only be measured along with the strategy.

The guiding criterion of an enterprise business intelligence system should be, "Does this system help our company achieve its strategy?" To answer this question, several intermediary questions must be asked. First, What is your strategy? For instance, become the market leader in a given segment, be number one or number two in all markets the company operates in, increase revenues by 20 percent, or increase market share by 30 percent. Next, What are your objectives set to fulfill that strategy? Divest from nonleading activities, improve customer service and customer retention, improve the bottom line, decrease costs in the IT department by 10 percent, improve customer retention rates by 5 percent, or reduce procurement costs. Lead a successful business reengineering program to transform the company into an e-business. The more precise these objectives1 are and the more they actually define and support the overall strategy, the more effective the related business intelligence system will be in helping achieve that strategy.

Third, What are the key performance indicators your company should track to measure success in meeting the objectives? A balanced scorecard initiative may help your company develop these metrics, though you need not implement a full-fledged, corporatewide, balanced-scorecard approach to start getting the benefits of business intelligence. Even starting with a handful of well-rounded indicators that are tracked only at the first few levels of an organization can put you on the right track. They must be agreed upon at the highest level and shared with everyone in the organization. For instance, if the company has strong goals in learning and growing, it may focus on how well employees are financially motivated to be creative. And finally, How can you use enterprise business intelligence to meet those objectives? Improve business efficiency by identifying opportunities or waste, access information that was not easily available, improve customer service, measure the progress made in reaching your objectives, take corrective measures as soon as you start deviating from goals.

BOC Gases
BOC Gases, a multinational supplier of industrial gases based in the U.K., decided against immediate implementation of a full-fledged data warehouse, after ERP applications from the German software giant SAP to run its core transaction processing took two and a half years to implement. BOC did not want to undertake a large warehousing project that could tie up nearly as many resources as the extensive SAP implementation.

Instead, BOC Gases opted for a flexible enterprise business intelligence system that would pull data directly from SAP applications. The implementation of a user-friendly enterprise business intelligence tool would insulate users from the complexity of SAP applications, avoid maintaining a staff of programmers adept at SAP's proprietary ABAP language to generate reports for the business staff, and provide business users with self-service access to the information they needed.

BOC Gases concentrated first on delivering information from SAP's sales data application. An enterprise business intelligence tool was provided to about 350 sales and marketing people in Europe. Through laptop computers, the group gained access to product and customer information specific to his and her areas of responsibility and are now able to drill into the data for details. The acceleration of business processes was substantial. Whereas previously they had to wait weeks after month's end prior to receiving paper-based reports, they now had easy-to-use information available electronically in a couple of days.

After the sales data implementation, BOC Gases followed up with an additional system for materials management, using enterprise business intelligence to create a snapshot of inventory levels that fall below predefined thresholds. Inventory reports are run nightly against SAP and published on the company's intranet. Another implementation used exception reporting—a means of highlighting outliers—to keep tabs on key business processes and quickly addressed problem areas. For instance, BOC was able to determine whether an interplant transfer order issued at one site had been completed at the corresponding site.

Lessons learned

Contrary to conventional wisdom, it is possible to gain great benefits by implementing a basic business intelligence system directly on top of an ERP, bypassing the need for a data warehouse.

—Bernard Liautaud with Mark Hammond

Business managers should drive enterprise business intelligence
Business managers drive the success of enterprise business intelligence. As one detailed analysis showed, the companies that were most successful in implementing enterprise systems were those that, from the start, had "viewed them primarily in strategic and organizational terms," as opposed to focusing on technical aspects.2

Gopal Kapur, president of the Center for Project Management, a project management consultancy group in San Ramon, Calif., argues that an effective business sponsor is the single most influential ingredient in the recipe for success of any IT project, enterprise business intelligence included. According to Kapur's checklist, in order to champion the project, the project manager, and the team, an effective business sponsor should:

  • Empower the project manager with appropriate authority
  • Remove high hurdles and keep the team out of political minefields
  • Support the team in resolution of cross-functional policy issues
  • Formally manage the project scope
Business managers are also in an excellent position to provide input into the front-end business intelligence tool with which users will interact. Companies invest large sums in data warehouses, enterprise systems, or other forms of databases, only to encounter dissatisfaction among users, because the tool they use is too complex or too difficult to use. Achieving a return on the investment depends heavily on how the users "take" to the system.

If business users find that the enterprise business intelligence tool is adapted to their needs, and makes their life easier, they will make return on investment happen. This is why it is crucial for business managers and users to drive the initial design, implementation, and training of the enterprise business intelligence system.

Maximizing return on investment through implementation
To ensure maximum business utility, the enterprise business intelligence system must be carefully rolled out. In particular, the users must be involved in the planning phase, the system must be as easy as possible, the users must be trained thoroughly, and management commitment must be ensured.

Involve the users in the planning phase. Business users must be involved from the beginning, in the planning and design phases, to ensure the system responds to all of their needs. It is also critical to involve users in the pilot phase to ensure that the terminology and standard reports are meaningful.

Allegiance Healthcare took care with the design and implementation of its ASPIRE system. For each component of ASPIRE—sales analysis, deal pricing, and customer proposals—groups of 10 to 12 users participated in the development effort. These users participated in requirements gathering, user-acceptance testing, and pilot rollouts. During construction, users were shown prototypes or up-to-date progress reports, to make sure the developers were meeting expectations. The user groups helped refine the customer proposal component of ASPIRE by showing the development team formats of proposals that had been in use.

Each of the three ASPIRE deliverables involved user acceptance, as well as pilot phases that were conducted in similar fashion. After receiving the software in CD-ROM format, conference calls were held with the user groups. The pilot phases also gave the IT development team a chance to better understand the capacities of the communication networks and database servers.

Planning was the only way to successfully accomplish a project the size of ASPIRE, says [Mark] Ciekutis [Allegiance's data warehouse manager]. Each development team created project plans, which flowed into higher level group plans. These plans then were rolled into an even higher level plan for the larger Horizon Project. Progress-to-plan was measured weekly, with progress, as well as spending, measured at the manager level monthly. Weekly team meetings were held at all levels, from senior management through subteam levels.

Make the system as easy to use as possible. The most successful enterprise business intelligence systems allow users to readily access data and to perform analysis with a graphical user interface that is relatively nontechnical, easy to use, and easy to understand. The number one success criterion for the implementation of a business intelligence system, mentioned over and over again, is ease of use. The ease of use of the system will also depend on how meaningful the data is for the business user. Ensure that the terminology used by the system corresponds to the business terminology of the business users. Finally, make the data as simple as possible.

The Belgian telephone company Belgacom took pains to ensure that its users were insulated from the complexity of the data warehouse to be accessed through business intelligence tools. "If top executives have to work with complex tools, they will never use them," says Koen Vermeulen, director of IT business analysis.

Train the users. While the best systems today are extremely user-friendly and intuitive, enterprise business intelligence is a new way of thinking; regular, customized training, based on users' own business problems, will help them extract maximum benefits from their data.

Most of the training will need to revolve around the data itself. Training on the use of software tools should take no more than a day or so. The critical path lies in educating the users about the data. What data is available? How is it categorized? What do the various business terms, such as customer and monthly sales revenue, mean? How often is the data updated? Data training should last about two days—understanding your data is about twice as difficult as understanding the tool.

Ensure management commitment. The importance of management commitment is critical. Because enterprise business intelligence is a new way of doing business, management has to lead the way. The introduction of such a system can be a significant departure from traditional business methods and may challenge conventional thinking. To battle common misperceptions, the highest level of management will need to sponsor the initiative.

One car manufacturer implemented an enterprise business intelligence system to help guide the allocation of funding based on the profitability of a project. The factory director made a rule whereby any project manager attending management meetings without a summary report generated by the system would not be given a chance to defend the project. This ensured that the data was kept up-to-date. In this instance, it took managers less than a month to become regular users of the tool.

Novartis Crop Protection
Formed by the merger of two major pharmaceutical companies—Ciba-Geigy and Sandoz—Novartis is one of the world's leading life sciences companies. Novartis Crop Protection is one of the companies in the Agribusiness division. Novartis Crop Protection manufactures a broad range of agricultural chemicals and has manufacturing facilities and sales units throughout the world. Crop Protection's commitment to sustainable agriculture focuses on the research and development of products to control the weeds, pests, and diseases that harm and reduce harvests.

One of the primary challenges of producing agricultural chemicals is that the production planning is done with a five-year lead time. This implies the integration and analysis of a great deal of external information, such as long-term market and sales forecasts. In addition, there are seasonal cycles that have to be managed. Most sales are realized from December to April. However, production has to be maintained at a constant rate throughout the year. Whereas the long-term planning is based on market forecasts, short-term planning is based on actual sales data from the group companies. Also, the management of production and distribution processes is complex, because it implies the combination of multiple materials and substances and involves numerous suppliers. Because of these constraints, having control over all relevant data is critical to Novartis Crop Protection success.

The company had standardized on SAP R/3 for both its finance and its supply chains, implementing many modules including Materials Management (MM), Production Planning (PP), as well as Warehouse Management (WM). The company decided to implement a data warehouse and standardize on a single business intelligence environment to access data across all data sources within SAP, and therefore cover most facets of running an agricultural chemicals company, including purchasing, production planning, sales and distribution, materials management, and finance.

One of the principal requirements was that the business intelligence solution had to be useable by anyone in the organization, regardless of position or technical skill level. It had to be able to do cross-process reporting, i.e., reporting across several modules of SAP. The value of data increases as users from different departments can access the same information, since they can then start cross analyzing that information with data from their own business area. At Novartis, the business intelligence implementation also had to isolate users from the complexity of the underlying SAP R/3 system. The tool had to be open to all sources of corporate data, including both SAP and non-SAP data. An additional requirement was to be able to produce a wide variety of reports from simple, ad hoc reports up to more elaborate EIS-style reports.

Because the implementation of a business intelligence solution was made in parallel with the SAP implementation (as opposed to being an afterthought), Novartis was able to implement the system in a short amount of time. Then, once the system was implemented, users throughout the enterprise were able to access data across all SAP modules and get much faster results than in the past. Enterprise business intelligence had compressed the reporting time from weeks to days.

Lessons learned

  1. Implementing a single business intelligence environment through a data warehouse allows access to data coming from multiple business processes at the same time, and consequently results in better understanding of the overall business.
  2. In the case of a complex system, consider starting the implementation of the business intelligence at the same time as the implementation of the ERP.

—Bernard Liautaud with Mark Hammond

Implement a "federated business intelligence systems" approach
A first look at enterprise business intelligence benefits may lead company executives to demand IT to create a centralized corporate data warehouse that incorporates all data needed at all levels of the company. The promise of a pure architecture and a single version of the truth may sound very attractive, but is highly unrealistic, and experience has shown that this approach will most likely fail. Integrating all data at all levels into a single environment is too formidable a task. Its implementation is too long and too costly. A preferred implementation is to accept the fact that many smaller business intelligence environments, based on many smaller data marts, will be built throughout the organization. By being smaller and closer to the data, and to the users who are the most involved in building and interacting with that data, these business intelligence implementations tend to be successful quickly.

Although these systems are not fully integrated, if built with an open technical foundation, they can be linked effectively. When users of other business areas can access the business intelligence environment of another business area, hidden facts are discovered, and consequently, the value is significantly increased.

A company implementing a federated approach can enable these disparate business intelligent environments to share common corporate metrics while letting them have their own local measures. Such an approach also trains the company on how to deal successfully with the unavoidable future addition of other back-end systems. As the company progresses in its information systems strategy, more applications are built to automate certain business processes, and more data warehouses are built on top of these applications. As market pressures increase, companies consolidate through mergers and acquisitions at a rapid rate. These acquisitions often force companies to embrace the business applications and data warehouses of the acquired company. A corporation architected around a federation of data warehouse systems that is used to efficiently open and link these systems can much more easily integrate to the new environment.

However, in order for such an approach to succeed, it is imperative that the federated systems be built around a standardized business intelligence solution. The product set must be open to embrace data coming from many databases and applications. Standardizing around one solution across the enterprise will enable users to communicate easily, to navigate across many departmental systems from the same system, and to adapt to changes, reorganization, or business model changes.

Fiat Auto's enterprise business intelligence, a Web-based system with 600 users and growing, is a great example of the federated approach. The main data warehouse serves as the hub for a network of interconnected data marts specific to business areas, such as finance, engineering, production, purchasing, human resources, sales, marketing, and others. "It has already paid for itself," according to Fiat CIO Castelli. Fiat Auto has spent about $10 million on the warehouse, but the company has already realized some $45 million in returns, largely due to cost comparisons that led to getting lower prices on parts.

As a summary, the federated business intelligence approach is more viable than any other for the following reasons:

  1. Each of the smaller environments has a much better chance to succeed.
  2. Incremental linkages are feasible and enable a progressive and less risky road toward building an enterprise view of the company's information.
  3. Departments feel more empowered.
  4. Companies are ready for consolidation.
As Douglas Hackney said in his January 2000 column, "The Federated Future," in Data Management Review: "A federated environment ... does not try to swim against the Tsunami of market and business forces driving nonarchitected/nonintegrated systems. Instead, it facilitates the integration of these systems, thereby avoiding the certain political death of opposing a powerful executive's tactical agenda."

The widespread use of information technology generates tremendous amounts of data. This data contains information that is invaluable to decision makers throughout an organization. The issue for most businesses is that the data is inaccessible to all but the IT department or the executives through complex systems. While most IT departments can run queries and produce reports at the request of business users, a self-service approach will provide the greatest benefits. Armed with precise, up-to-the-minute information, users can develop effective responses that help attain company goals.

The key to leveraging this wealth of data is to implement an enterprisewide business intelligence strategy. It can reduce costs, increase revenues, improve customer satisfaction, and improve cross-company dialogue. While many of these benefits are clearly quantifiable, some of the more intangible ones—such as improved communication, improved job satisfaction, and sharing of intellectual capital—will give the greatest edge over the competition.


  1. Large organizations tend to define their strategy, then related goals (nonquantifiable targets that support the strategy), then objectives that define each goal. Smaller organizations tend to define the strategy and then directly define objectives. In either case, objectives must be S.M.A.R.T. (specific, measurable, achievable, relevant, and time bound).
  2. Thomas H. Davenport, "Putting the Enterprise into the Enterprise System," Harvard Business Review (Cambridge), July–August 1998.