Distributed computing is forcing EIS to rethink its traditional host-based ways
Karen Watterson
It used to be that everyone understood what you meant when you said "EIS" (Executive Information System). An EIS was an electronic briefing book that summarized key sales or production figures for busy executives. Early EISes tended to be custom-built by MIS staffers assigned to top-level executives.
As organizations implement more distributed client/server networks, however, decision-support applications like EIS and the related DSS (Decision Support System) have been forced to adapt. Both end-user data-access tools and client/server development tools need to analyze larger amounts of fast-changing data. This is especially true in transaction-intensive businesses like retailing and finance.
And EIS/DSS must now support more than just a few executives.
Organizations spend millions of dollars collecting, storing, and safeguarding data. The challenge for today's EISes is to make that data accessible and usable to the people on the network. That means putting the tools to manipulate the data into the users' hands so they can answer their questions and understand their markets better. They don't want to go through the IS department. IS, according to Jeffrey P. Stamen, president of IRI Software (Waltham, MA), is often better being the "system police" than understanding the needs of sales and marketing departments.
The trick is to present just the information required in just the right format without rekeying summary data from massive printouts into spreadsheets. The good news is that today literally hundreds of client/server products provide reporting and querying capabilities, often at a fraction of the cost of classic host-based EISes.
But the classic EIS and DSS vendors aren't
standing still. They're adopting client/server architecture, becoming more open and modular, and adding value by targeting vertical markets with special versions of their products. Most emphasize their ability to work with mammoth databases, their understanding of MIS, and their track record.
How EIS Works
EISes are generally designed with two or three parts: an administrative module, where data access is managed; a builder module, where a developer sets up data mapping and builds a sequence of screens; and a run-time module that the executive or knowledge worker runs. Sometimes the first two functions are combined.
Data access and consolidation vary depending on the package. Some EISes provide their own data storage system; some only package the data and route it into a more accessible database--usually on a LAN. Some EISes do the data selection and consolidation on the host; others do it at the workstation. Some use third-party commercial middleware gateways or popular APIs like ODBC (Open
Database Connectivity); others are less open, offering only a discrete list of proprietary drivers--and often at extra cost. ODBC drivers, on the other hand, are widely available, often bundled with other products. Finally, almost all of today's EISes come with a standard GUI, such as Windows, Macintosh, Motif, or Presentation Manager.
Robin McNeill, director of PowerPlay products for Cognos (Ottawa, Ontario, Canada), identifies six tasks that managers do for which an EIS will be useful: track, flag exceptions, rank, compare, trend-spot, and investigate and explore. The features most EIS tools provide reflect these tasks. They present summarized and consolidated data in both report and chart form, or they allow sequencing of screens to produce executive slide shows. Hot spots that users can click on to get more information and other drill-down techniques help users navigate through varying levels of detail.
Exception reporting is another extremely useful technique managers use to flag data that'
s unusual or out of bounds. Both unusual and periodic events can be defined to trigger visual or audible alarms or activate an intelligent agent to perform a specific task. Some EIS packages support statistical techniques like regression and correlation analysis, which let you measure the strength of relationships between pairs of variables. Others include a spreadsheet interface, ad hoc queries, and batch processing of reports.
The tasks that EISes track are often mission-critical. For example, Chevron's IT subsidiary built a sophisticated $750,000 inventory-tracking system for Chevron U.S.A. Products that helps optimize the oil-refining process and minimize idle inventory. Part of that system uses Cognos's PowerPlay to show how oil moved through the supply chain during a given production week. Don Waddell, a Houston-based project manager for inventory coordination with Chevron U.S.A. Products, says, "We found the best approach for decision-support systems was to have all necessary information availab
le on a summary basis in a consolidated system." Waddell's group used Oracle to create a data warehouse of consolidated data from diverse mainframe data sources, including IMS, Nomad, and VSAM files across an SNA (Systems Network Architecture) network.
In large law firms, time and billing applications are mission-critical. Bob Warrick, senior programmer/analyst with Pillsbury Madison & Sutro, a large San Franciscobased law firm, has built a PowerPlay 4.0 application that lets the executive group of attorneys track billing performance by practice group, location (the firm has 12 offices), and type of attorney. Time and billing information is migrated from a PowerHouse application residing on a VAX cluster to a PowerPlay database and retrieved via DECnet.
The Changing Face of EIS
Your choices for decision-support tools are confusing. Low-end (read "inexpensive") data-access and client/server tools and spreadsheets mimic much of the functionality of the classic EIS. Data warehouse and replicati
on packages provide "safe" copies of nonoperational data that users can manipulate to their hearts' content. Smart middleware, such as InfoPump from Trinzic (Palo Alto, CA) or Micro Decisionware's Database Gateways (Boulder, CO), can be programmed to periodically download information to refresh the data sets. Client/server packages with intelligent agents can automate processes such as warning a financial analyst when a key ratio has been exceeded or sending E-mail to a purchasing manager when an inventory-reorder quantity has been reached.
Many organizations are meeting the challenge of providing direct access to corporate data on the mainframe by creating LAN-based data warehouses that contain read-only snapshots of host data that's periodically refreshed. This has the advantage of minimizing network traffic, expensive host CPU time, and security headaches. It also provides nonintrusive access to data in legacy systems. Some people refer to this approach as a client/server/server system or a three-ti
er architecture. Sometimes the data warehouse is referred to as a datamart or staging server. Data warehouses often contain data from multiple data sources that has been consolidated and summarized. This is not on-line production data; it might be updated daily or weekly as needed. But it's real-enough time for most decision making.
Vendors such as Red Brick Systems (Los Gatos, CA) offer a different vision of a data warehouse. Red Brick Warehouse 2.1 ($20,000 and up) is an RDBMS (relational database management system) that's optimized for queries rather than data entry. Typical Red Brick clients are firms that need to analyze large amounts of data, like retailers and financial institutions. Red Brick relies on lots of indexes to allow fast retrieval of databases that average over 10 GB of data. Indexes are special files that work like indexes in a book. Instead of pointing to a page, they point to the exact location of a database record, based on the index field. If you index on customer number, compan
y name, contact name, city, and area code, for example, retrieving information based on any of those fields is faster than it would be without indexes. Because of the overhead associated with maintaining the index, on-line production databases can't afford to index on all fields; data warehouses can.
Another firm, Dynamic Information Systems Corp., based in Boulder, Colorado, sells a product called Omnidex that generates indexes in your current databases (in contrast, Red Brick creates a new database). DISC also provides an API that developers can call from Windows front-end tools. Chris Werle, a business systems analyst, uses Omnidex to provide managers at Chicago-based Boots Pharmaceuticals with up-to-the minute sales and order-entry information. He likes the fact that Omnidex reduces network traffic. "Sometimes," he says, "we just want to know how many there are--to do the equivalent of a SQL COUNT. We don't need to see the records themselves at all." The project's success has prompted Boots to begi
n bringing previously outsourced data processing back in-house to build another EIS for marketing and sales.
Prism Solutions (Sunnyvale, CA) offers yet another angle on warehousing. Prism Warehouse Manager generates COBOL, SQL, and JCL or Script code to move legacy IDMS, ISAM, VSAM, IMS/VS, and COBOL data files into Oracle, SQL Server, Red Brick Warehouse, NonStop SQL, Teradata, or DB2 databases. Trinzic's InfoPump takes yet another approach, offering intelligent middleware for routing, integrating, and synchronizing dissimilar data. These warehousing, indexing, and routing products package and consolidate data that's often dispersed among multiple data sources. Ultimately, this minimizes network traffic and makes it easier for the EIS and other software to access the data.
Replication servers are also becoming more widespread. Replication is related to the notion of data warehousing, but data warehouses usually contain only a subset of the data, while replicas are usually copies of an entire
database. Sybase, Ingres, Informix, and Oracle sell replication servers that can keep multiple copies of the database up to date, making it easier for users who may be physically dispersed to get fast, local access to data. The industry can thank Lotus Notes for popularizing the concept of replication, since replication of Notes databases is a common activity. Like warehousing, replication makes copies of data conveniently available for access by EISes.
Another force affecting traditional EISes is that, more often than not, today's executives are computer literate. Many of them are familiar with spreadsheets, drill-down techniques, and slice-and-dice techniques for data exploration. They may not want predigested summaries.
Even more significant, though, is the trend toward flatter organizations with distributed decision making. It's no longer just executives who need access to corporate data, and, in a sense, EIS as such is really a dinosaur. Line and branch managers need sales and marketing dat
a, and customer-service representatives need access to customer account and credit information. Classic EIS let you build static presentations for upper management. It wasn't designed for interactive, ad hoc exploration of the data. If it gave that illusion, it was thanks to the skill of the programmer who anticipated the executive's needs. "The flattening of management structures," says Frederick Lizza, vice president and general manager of Trinzic's Database Access and Connectivity Unit, "makes the limited hierarchical executive information system a relic of the past."
EIS is also feeling price pressure from spreadsheets, which have data access built into them. Spreadsheets have a wide installed base and are the tool of choice for millions of business information workers and decision makers. These users don't want to have to learn another program to do their data analysis. As spreadsheet vendors add more functionality, like pivot tables and multidimensional spreadsheets, EIS products will come under
even more pressure. Multidimensional spreadsheets make it easy to answer questions that would be extremely cumbersome in SQL, such as "Give me the top three sales regions based on the percentage change in revenues this year relative to last year."
Associated with multidimensional spreadsheets is a trend that's sometimes called OLAP (on-line analytical processing). In contrast to OLTP (on-line transaction processing), which focuses on order entry and other transaction-processing systems, OLAP includes decision-making systems for marketing, sales, and finance (see the text box "The 12 Rules of OLAP" on page 184). As IRI's Stamen explains, "OLAP is a way of looking beyond transactions to the forces driving them. It can help companies accurately forecast sales in order to better plan inventory and production levels, know where advertising is working and where they're wasting millions of dollars, and determine if their products are correctly priced."
In addition to mainstream client/server and sprea
dsheet programs, the EIS market is being invaded by special-purpose packages, including statistics packages. SAS (Cary, NC), for example, has a SAS/EIS module. Data-discovery software--e.g., IDIS (Information Discovery System) from Los Angelesbased IntelligenceWare--uses statistical methods to find correlations in data.
The Depository Trust Company is one of the world's largest clearinghouses for the securities industry. Company executives set goals and determine tasks required to meet those goals; managers must then provide periodic status reports so that senior management can monitor progress. SAS/EIS performs this task well because it provides easy access to the firm's DB2 data without duplicating it, according to Khasha Dehnad, manager of decision support and executive information.
Other applications are performing EIS-like functions. Multidimensional databases such as Kenan Technologies' Acumate (Cambridge, MA) store data acquired from external data sources much as data warehouses do. Howev
er, multidimensional databases (not to be confused with multidimensional spreadsheets, although both typically include features that support financial modeling and statistical analysis) let developers define multidimensional "views." (RDBMSes are limited to 2-D views.) EIS-like functions are also apparent in natural-language software. With Natural Language from Natural Language, Inc. (Alameda, CA), for example, once the developer defines the interface, managers can formulate questions in English (supplemented by words in other languages, if desired) rather than in SQL (Structured Query Language, a sort of Esperanto for relational databases).
Add these to the traditional DSS and high-end financial analytical modeling packages, and you've got the makings of a highly competitive--and confusing--market. "DSS and EIS systems are on a collision path with traditional reporting and workstation tools gaining EIS-like graphical features," says Ron Sella, business development manager of the Microsoft Products Div
ision of Information Build-ers (New York, NY); "DSS and EIS will merge within the next two to three years." Sella points out that Information Builders' new Focus Reporter for Windows, for example, already incorporates EIS features such as drill down.
Michael Saylor, president and CEO of MicroStrategy (Wilmington, DE), which makes EISToolKit and DSS Agent 1.0 Enterprise, takes a more Windows-centric view of the market. "Up until 1993, the EIS/DSS market consisted of maybe 30 or 40 players," he says. "Imagine them playing volleyball on the beach. Then along came Bill Gates with this huge bulldozer representing Windows, complete with its APIs and Visual Basic, which essentially commodified a huge amount of EIS. This giant bulldozer was plowing along, scraping a foot of sand off the beach. Some people ran headlong into the bulldozer. Others ignored it and were plowed under. Some tried to escape into the ocean. Others tried to outrun it. Our approach is to look for the opportunities behind the bulldozer. Th
ere's a tremendous need for vertical-market applications and other new kinds of tools. Why fight the bulldozer?"
Consultants like Mark Burgess, president of San Diegobased Knowledge Works, would agree. Instead of using a commercial EIS, Burgess opted to use Visual Basic as the engine for a prototype budget-analysis EIS he built for the U.S. Air Force. Burgess's EIS application runs over a WAN (wide-area network) to access distributed data, which will probably reside in a SQL Server for NT database, using Microsoft Access for staged data.
Common Themes
All the applications competing to perform EIS tasks have some common themes. They are being used by more than one person. They use GUIs, most often Windows. Many automate what amounts to complex SQL queries as well as data consolidation and charting in a client/server framework. And that's the rub. EIS and DSS are merging, to be sure, but the bloodier battlefield will be where client/server meets EIS/DSS.
Today there are too many players,
and that's a formula for industry consolidation and shakeout. Expect to see most EIS vendors shedding the EIS moniker, with new releases of their products emphasizing enterprise, client/server, multidimensional, business information systems, or analytical processing. Expect increasingly open and modular systems. Proprietary EIS systems are surely dying, but the death rattle may be a prolonged one, similar to that of COBOL and mainframes.
You can also expect increasing numbers of both classic and new EIS vendors to target special vertical markets. For example, IRI Software already has special software for the consumer packaged-goods market, where its parent company Information Resources does market research. Farther down the pike, you're likely to see intelligent agents that perform processes based on predefined conditions. The logical next step will be to extend automated analysis to automated decision making, according to MicroStrategy's Saylor. After that, you should see what Saylor calls cybernetic
decision making, where only the best intelligent agents survive in a self-modifying system that is constantly perturbing and sampling its own systems.
Dimensions of EIS
The capabilities, implementation, and use of EIS vary widely. The extremes to which this is true are illustrated below.
Single user Collaborative
Type of report Canned Interactive, supports
drill down
Data analysis Internal only Competitive
analyis
Corporate planning Tactical Strategic
Cost Expensive in terms Cheap enough for
of initial cost and deployment on
consulting and the desktop
licensing fees
Architecture Stand-alone data- Integrated, part of
analysis tool multicomponent
package
Compatibility Proprietary Open and modular
Illustration: The Decision-Support Pyramid
As EISes try to adapt to the client/server world by performing more tasks, other products are encroaching on the traditional EIS turf.
Illustration: The Acutrieve component of Kenan Technologies' Acumate Enterprise Solution acts as a gateway to the company's Multiway multidimensional database engine.
-- Icons for tables, graphs, drill downs, rotations, subsetting,
data browsing, and analysis
-- Data displayed in multidimensional tabular format
-- Analysis copilots are on-screen helpers that guide the user
through analyses such as linear regression or forecasting.
-- Graphs and tables can be linked to display the same data or
made independent.
Karen Watterson is the principal of the Watterson
Database Group, a consulting firm based in San Diego, California. You can reach her on the Internet
at
111-9390@mcimail.com
or on BIX c/o "editors."
The Changing World of EIS
