BI today: One version of the truth
- By Jack Vaughan, Peter Bochner
|Danny Siegel, (pictured at left)
senior manager of business technology at Pfizer Inc.’s Global
Pharmaceuticals Division in New York City, notes, an effective BI system
provides corporations with “one version of the truth” that can come from
developing a system that provides consistent or harmonized numbers from
all corporate divisions. |
Business intelligence (BI) describes the process firms go through to gather, store and analyze data. It is a critical activity that helps companies to make faster, smarter decisions, as well as increase revenue, build customer loyalty, streamline operations, improve risk management and even enable previously impossible business processes.
Historically, BI has been hard to define, since it is neither a product nor a system. Shaku Atre, president of Atre Group Inc. in Santa Cruz, Calif., calls BI “a constantly evolving strategy, vision and architecture that continuously seeks to align an organization’s operations and direction with its strategic business goals.”
Or as Danny Siegel, senior manager of business technology at Pfizer Inc.’s Global Pharmaceuticals Division in New York City, notes, an effective BI system provides corporations with “one version of the truth” that can come from developing a system that provides consistent or harmonized numbers from all corporate divisions. That task is especially difficult at companies like giant drugmaker Pfizer, which operates divisions worldwide and is constantly on the lookout for expansion via rollups and mergers.
But as hard as it is to define BI, it has been harder to implement. “More than half of all BI projects are either never completed or fail to deliver features and benefits that are optimistically agreed on at their outset,” said Atre Group’s Atre. There are many reasons for this failure rate -- high cost of ownership, lack of ease of use, organizational issues, lack of measurable benefits, benefits restricted to few users, a lack of scalability and so on.
Costs have come down, and many think ease of use is not an obstacle now that dashboard and other visualization technologies are helping users to understand trends in their data. “We thought making graphics interactive would appeal to the higher-end analyst,” said Michael Corcoran, chief communications officer at Information Builders Inc., New York City. “Instead, novice users are finding it an intuitive way to navigate the data.”
But organizational issues persist. “Getting all the various business units, who often have different roles, to buy into the same initiative is the number one obstacle,” said Ken Chow, VP of marketing and product management at Group 1 Software, Lanham, Md.
Interoperability issues are also starting to fade. More companies are looking for a pre-packaged, well-integrated stack that does end-to-end BI. “Users are getting tired of BI solutions that don’t interoperate,” said Paul Turner, director of products for Sunnyvale, Calif.-based Hyperion Solutions’ BI platform. “But we’re starting to see businesses standardize on all their BI initiatives.”
There is one thing everyone agrees on: A deluge of data is impeding BI. “The theory is that the more you know about your customers and the business problem you’re trying to solve, the better you’re able to solve it,” said Karen Parrish, VP, worldwide sales, BI solutions at IBM. But by trying to access data from too many sources -- data that resides in their own organization, data that resides externally, data that they purchase and bring into the organization, data from the Web and data that sits in e-mail -- companies may be shooting themselves in the foot.
“People today are collecting all this data, but they can’t get at it in a meaningful way,” said Craig Schiff, president and CEO at BPM Partners in Stamford, Conn.
Requirements for corporate compliance, such as HIPAA, Sarbanes-Oxley and Basel II, are forcing firms to keep data longer. Basel II, which is about the management of cash portfolios and risk determination, requires financial institutions to keep three to seven years of data on individuals they have relationships with.
The larger and more complex the data, the harder it is for companies to conduct BI. “And the problem is going to get worse,” said Jit Saxena, CEO at Netezza in Framingham, Mass. This difficulty in carrying out BI when the amount and complexity of data grows too great is preventing telecom carriers from analyzing billions of call detail records to capture fraud and cross-carrier billing. It is stalling data mining on customer records. And it is slowing research in areas like bio-informatics.
Pfizer seeks consistent data
When financial managers meet, they come armed with data from modern BI systems. But one problem arises: Derived as they are from desktop spreadsheets, the numbers of one person do not always jibe with those of another. And that is even after knowledge workers have spent time finding and massaging corporate data to get to the truth.
The proliferation of data and data warehouses is partly to blame, as are legacy, client-centric BI systems.
Bringing “harmonized” data to the decision-making process is a major effort today, as Pfizer Inc.’s Siegel can attest.
Pfizer is the maker of such prescription drugs as Lipitor, Viagra and Zyrtec, among others. According to Siegel, “a lot of financial folks were spending the majority of their time collecting and integrating various data,” when the group asked him to pursue the idea of integrating several diverse financial systems a few years ago.
Mergers and rollups have marked the pharmaceutical industry in recent years, each bringing more work for IT in the form of legacy system integration. One of the most notable mergers in the pharmaceutical industry was Pfizer’s acquisition of Warner Lambert in 2000. Like others, Pfizer’s IT pros must also wrestle with the results of smaller acquisitions.
The important thing is to ensure that “the numbers” are consistent, said Siegel. But consistency can be difficult, as different groups may have various ways of measuring things. Moreover, the elements combined to arrive at a financial line item can change from spreadsheet to spreadsheet. The truth, in that light, is fluid.
For example, what is the right answer to a simple question such as “What were the sales of [a particular drug] in Germany last year?” The answer, said Siegel, is: “It depends on who you ask, when you ask it, how they mapped their data, or how they defined something like profit and loss.”
If a reporting group uses a global calendar, or includes cross-divisional charges, truthful answers to the same question can vary. In a presentation at The Data Warehouse Institute (TDWI) conference in Boston in May, Siegel asserted that financial reporting in complex organizations has many faces. Terms can have multiple meanings.
The result is that knowledge workers can have a meeting, track an accounting question and have spreadsheets with different financial numbers. What is needed, said Siegel, is “one version of the truth.”
Technology is a means to an end for Siegel. “Data warehouses are OK, but the trick is that you need to build data standards around the data warehouse to make it meaningful,” he noted.
Understanding the need for data standards has been key to his efforts at Pfizer. The firm has used Ascential Real-Time Integration software in its effort to streamline gathering BI around data standards. “You could call it meta data,” Siegel said, “but ‘data standards’ is a more useful term.”
The data warehouse industry as a whole has made progress, but vendors sometimes still “skip a step,” he said. “The tool is only as good as the base below it. It does not work if things are not based on sound business logic, good hierarchies and good data. A lot of meta data maintenance tools still have a technology point of view.”
The ongoing effort by big companies to conform to Sarbanes-Oxley governance requirements may help managers to better focus on the business view. Referring to Sarbanes-Oxley, Siegel said that once you have documented how your business runs, then you can better address the question of what you want to do with the tools and technology you have.
Medical testing firm tries BI
Molecular Pathology Laboratory Network Inc., Maryville, Tenn., was founded to apply molecular techniques to diagnostic testing for cancers and infectious diseases. Four years ago, the amount of data the lab had to manage skyrocketed when the number of specialized tests it performs doubled to about 5,000 over a period of 18 months.
“Drowning in data was a fair description of where we were,” said Allen Hunter, who heads up software development at Molecular. At that point, the lab realized it needed to overhaul its processes for handling reporting and data analysis, which consisted of multiple legacy applications.
The firm decided to build in-house a Web-based Laboratory Information System (LIS) that could scale to handle growth of the number and types of tests done as well as the number of clients. LIS runs on an HP server running Windows 2003 and is built around the Cache post-relational DBMS from InterSystems Corp., Cambridge, Mass., which can support real-time data analytics through the use of bitmap indexing.
The Web-based LIS feeds data from lab instruments and devices into the database, giving internal users fast, easy access to tens of thousands of diagnostic test results. The system also includes complete client reporting and client billing capabilities. Currently, the core database for the LIS is about 1.5 Gb and includes information on tests for an estimated 42,000 patients.
“We’re handling around 7,000 tests each month, and we expect continued growth in demand for data,” said Hunter.
So far, handling all the data needed has not required any special storage technology but, as Hunter noted, with the distributed database capabilities that Cache provides, “we can easily add extra hard drives and servers on an as-required basis.” As data continues to grow, he said, “we might have to go with a separate archived data warehouse, but as long as we can keep adding hardware, we don’t foresee any major problems.”
Motoring to the right data
Not only are companies drowning in data, but many of them are drowning in the wrong data. “A lot of the companies I’ve visited recently are sourcing 20 different data feeds,” said Hyperion’s Turner. “They have ERP systems, spreadsheets, text files and a data warehouse. It’s unreal how much stuff they had to integrate, and it was all in different formats with no common view of customers.”
Data often does not flow properly across the enterprise. It is difficult to get to “a single version of the truth” when people across departments and regions are dealing with different information in silos.
“More companies are seeing the value of integrated reporting, modeling, planning and analysis” said Netezza’s Saxena. “The silo approach to BI, with every department having its own data and BI, has limited appeal.”
American Suzuki Motor, the U.S. subsidiary of Japan’s Suzuki Motor in Brea, Calif., markets Suzuki cars, motorcycles, ATVs and marine engines through a network of 420 dealerships in 49 states. The company’s strategy for increasing market share is to improve customer satisfaction by raising dealer performance.
The plan involved storing sales and warranty claim data in a database. District managers would use this data to create reports on dealer performance, and then coach the poorer performing dealers on how to improve. At first, the Suzuki organization kept raw statistics, but they were not correlated and therefore were not analyzable. Four years ago, the company realized the approach was not viable.
“We needed a BI product,” said Todd Bowers, systems analyst for business intelligence at Suzuki.
A committee chose Hyperion’s Essbase multidimensional database and the Brio Intelligence data mining tool. (This was three years before Hyperion purchased Brio in October 2003.) Both products ran off an IBM I-series AS/400, with IBM’s DB2 as the DBMS for source data.
The source data flows from DB2 to Hyperion to Brio, although Brio can also retrieve data directly from DB2. Sales and warranty claim data is stored in Essbase, and its multidimensional structure allows users to drill down into that data.
Still, the system needed tweaking. The different vehicle groups each had their own models and their own way of grouping things. Nothing was centralized.
“The data load was high and, because the detail level was hard to work with, people were creating their own silos in the format they wanted to see the information,” said Bowers. “Users were interested in getting aggregated data but were getting bogged down in the details.”
So in early 2002, Suzuki brought in Claire Ashby to create a data warehouse to tie together the data from all the groups in a logical business model. “Today, we can take transactional data not only from sources but also from each department’s routing,” said Ashby. “This means each department can have their own views and grouping based on their own data.”
But, noted the firm’s Bowers, “We’re getting away from silos. I can look into Essbase the way I want to see something and someone else can see it in their way, but it’s the same data.”
BI serves appetizers
Nancy’s Specialty Foods, a producer of gourmet appetizers in Newark, Calif., makes more than 35 tons of food products each day. Like other food and beverage manufacturers, Nancy’s ability to control costs for the food it produces is impacted by fluctuating prices for things like dairy products, produce and eggs.
In 1997, Nancy’s moved its forecasting, budgeting and analyzing of sales information from Lotus and Excel spreadsheets to a deployment of Applix TM1 for BI. “Nancy [Mueller, company founder] was frustrated with her sales VPs coming in with five different sets of numbers each month,” said David Siegfried, manager of MIS at Nancy’s. “She wanted a
solution with more data integrity.”
The company uses the MFG/PRO ERP system from QAD and a Progress Software DBMS that runs under HP-UX. TM1 resides on a Windows 2000 server. “We consider the TM1 side a data warehouse because that’s where we do all our analytics, planning, forecasting and results reporting for the various areas of the company,” said Siegfried.
Bolstered by the improved understanding of product profitability it gained through BI, Nancy’s has rolled out a new BPM app based on TM1 Version 8.0. Nancy’s relies on the app for what-if scenarios to understand product costs with many variables, as well as to view the financial impact of its production schedule three to six months out.
By slicing and dicing the product and sales information, Nancy’s is able to better understand profitability by product to make informed decisions about expanding a line or discontinuing a product.
Please see the following related story: “A look at BI vendor strategies
” by Peter Bochner