Data Warehouse Economics Update (II): Opportunity Knocks

Issue Date: 01/01/2000
Issue Number: 10.08
Category: DATA WAREHOUSE

Two years ago, we surveyed a cross section of users on what was then a new technology, finding that many data warehouses were being lead forward by a 'build it and they will come philosophy.' At the time, few users had any solid idea on what they could actually do if given the right tools. With the technology having matured, the question is whether implementations are being driven by more 'normal' investment guidelines. In this report, we analyze the experiences of 10 users-including a return visit back to a member of our original study sample.

STUDY SAMPLE

As shown in Table 1, the projects examined for this report ranged in maturity, from brand new to a decade old. The sizes varied greatly as well, ranging from user populations of a few dozen to a few thousand, with the companies ranging from manufacturers to healthcare, consumer products, and distribution services.

Table 1. Research Samples
(Source: Computer Finance)

Customer	DW System	Status	User Population	Goal(s)
Boeing Employees	Sagent	1 mo.[1]	51	Marketing Analysis
Credit Union				Customer Demographic Analysis
Chicken of the Sea	Seagate Analysis, Info	3 mos.	30	Sales Analysis
				Operations Analysis
Dade Behring	Cognos PowerPlay, Impromptu	2 yrs.	300 - 400	Sales Analysis
	Cognos SAP Accelerator (embeds ActaMart)
Hallmark Cards	Microstrategy Suite	4 yrs.[2]	500 (current);	Sales Analysis
			1000 - 2000 (future)	Customer Analysis
				Product Analysis
Imperial Tobacco	Cognos PowerPlay, Impromptu	6 mos.	35 - 40	Accounts Receivable
				Billing
J.R. Simplot	Brio Reporting, Analysis	18 mos.[3]	250	Sales Analysis
				Supply Chain Management
Kaiser Southeast	Cognos PowerPlay	3 yrs.	150	Pricing Analysis
				Financial Analysis
				Care Delivery Quality
				Customer Reporting
Nissan North America	e.piphany	3 mos.	30	Sales Reporting
				Cross-Selling, Up-Selling
				Campaign Management
Owens & Minor	Business Objects WebIntelligence	6 mos.	5 product suppliers;	Supply Chain Management
	Informatica PowerMart		40+ healthcare system customers
Vlasic	Brio Query	1 yr.	200	Campaign Management
	Informatica PowerMart			Logistics Analysis
				Customer Service
				Sales History

 

What's not surprising is that sales and marketing data marts and warehouses by far predominated our sample-a result that was also quite typical of first-generation implementations.

Unlike transaction systems, business intelligence systems are optional. Therefore, when dealing with non mission-critical technology, the investments that have the most overt potential impact on the bottom line (e.g., boosting sales or profits) are likely to get approved first.

But we also saw a significant growth in the area of operations, with applications ranging from corporate finance to product development, service quality, and supply chain management. In the latter case, the organization was building the data warehouse as part of a new business, selling new value-added services to its trading partners.

What was also interesting was that corporate restructuring played a significant role; several organizations rated building a data warehouse as a way of leapfrogging the limitations of legacy systems that were being passed down to them. In some cases, they opted to build data warehouses before attempting new transaction systems. Out of our sample, three out of the ten organizations that we contacted found themselves in corporate restructuring situations, which at first glance would appear a freak of sampling. However, given the accelerating rate of M&A activity of the past few years, maybe our findings weren't so unusual.

We contacted the following organizations for this study:

Boeing Employees Credit Union. The third largest credit union in the U.S., serving 250,000 customers, the organization has state-of-the-art systems and a relatively sizable IT staff of 75 (out of overall staff of 900). The IT environment is highly mixed, consisting of UNIX and OpenVMS servers, and NT clients, and an application portfolio comprised of PeopleSoft financials, Lawson HR, and separate transaction systems for different banking products (e.g., savings, checking, mortgage, lending, etc.). As part of an initiative to introduce more electronic banking products, the credit union is building a data warehouse to target customers more effectively.

Chicken of the Sea. A privately-held 3000-person company known for its household brand canned tuna products, the dominant application is PRMS, an AS/400-based financials and MRP II system. The company installed the Seagate Info reporting system on an NT environment to provide ad hoc financial reporting; future plans call for inventory, sales, and forecasting reporting systems.

Dade Behring. One of the top three suppliers of medical diagnostic devices worldwide, the company is the product of corporate restructuring. First spun off as an independent company 5 years ago from Baxter, it subsequently acquired the diagnostics division of Dupont and the European-based diagnostics division of Hoechst.

Over the past two years, they built sales data marts, followed by inventory and finance using the Cognos PowerPlay OLAP system. It was done in parallel with SAP projects; the North American operation implemented R/3, while the acquired European unit already had a mixture of R/3 and R/2 (which will eventually migrate). Meanwhile, operations in smaller nations implemented Systems Union, a PC-based package. The data marts were seen as ways of bridging all the systems. They typically email about 350 PowerPlay OLAP cubes to individual recipients each month, but are looking for web alternatives because the size of the cubes is approaching up to 25 megabytes, zipped.

Hallmark Cards. The world's largest and best-known manufacturer of greeting cards, Hallmark built its first data warehouse in 1989, with a homegrown application running off a Teradata database. In 1995, it replaced the legacy warehouse application with what is now called the Microstrategy suite, and began migrating Teradata to the current UNIX platform. It did so because using an off-the-shelf product would in the long run make it easier to add new functionality.

Significantly, most of the application logic was in place prior to implementing Microstrategy; what's changed, however, is the sheer size of the Teradata data warehouse, which, over the past five years has mushroomed tenfold to roughly 5 billion rows, serving hundreds of users. With expansion to the web, the company hopes to reach up to 2000 users. The data warehouse covers everything from product sales (including point-of-sale transactions from its own stores) to nearly two dozen product features, and provides a mix of standard and ad hoc analysis capabilities.

Imperial Tobacco. Canada's largest tobacco company, the firm has three sales offices that deal with 200 distributors across Canada. Each day, the company processes a huge volume of accounts receivable. Although it recently implemented SAP, reporting required time-consuming ABAP 4 programming; when it considered alternatives, SAP's Business Information warehouse (BW) was not available. A year ago, it implemented Cognos tools (PowerPlay OLAP and Impromptu reporting) to replace its previous, labor-intensive manual reporting, and now delivers HTML reports to 35-40 users in the finance department.

J.R. Simplot Food Group. A 7500-employee diversified agribusiness business unit (which accounts for just over half the parent company), the challenge was to integrate data from multiple legacy systems into a format intelligible to knowledge workers. The company has eight major business segments, ten major data sources. Over the past five years, it has struggled piecing together a data warehousing strategy that could effectively consolidate information from multiple, often poorly-documented sources. Initially, it built systems using Sybase IQ and end-user Lotus Approach databases, but the complexity of the system (which required end users to do table joins for query processing) proved too complex. Two years ago, they standardized on Microsoft SQL Server and the Brio query and reporting tool, distributing prebuilt OLAP cubes to nearly 250 users.

Kaiser Permanente. When we last spoke with Kaiser's Atlanta-based Southeast division for our report two years ago, the unit was ramping up its recently-completed data warehouse, using Cognos PowerPlay OLAP cubes, which join data from membership, claims, clinical, pharmaceutical, and financial systems. At the time, PowerPlay was chosen because of its ability to consolidate data from multiple databases. With the system operational for over two years, the data warehouse has become a key part of the organization's business processes. However, end user acceptance remains an issue. Casual users often still rely on the power users to provide information that they could obtain themselves. While some line organizations have heavily embraced the system, in others usability issues proved obstacles. The original Windows client/server front end proved too complex for casual users, who would be better served with simplified, web-based managed reporting environments. Additionally, as transaction systems were upgraded in other parts of the organization, in many cases the architecture and content of some of the OLAP cubes had to be changed.

Nissan North America. Customer loyalty has proven a major marketing hurdle for the North American unit of the Japanese automaker. Although its cars have been consistently rated as offering better values than Honda or Toyota, Nissan was not getting as many repeat customers. The company has numerous standalone systems that each housed different aspects of customer data. 'We realized that we had to develop a system to study sales transactions, customer behavior, and demographics to help us identify real selling opportunities,' said Ted Ross, corporate manager of owner loyalty and database marketing.

The initial project was putting together an Oracle-based operational data store for converging customer data, then looking for CRM (customer relationship management) solutions to provide more focused views. Nissan chose e.piphany because existing CRM packages were deficient in campaign management (planning and evaluating the results of marketing campaigns) and lacked data mining capabilities that would allow them to dissect customer buying patterns.

Nissan bought two e.piphany modules: sales reporting and analysis, for analyzing customer activity; and cross-selling, to identify opportunities for selling additional or upgraded products to existing customers. It would draw data from fiver major sources: sales transactions, financing, and syndicated data listing registered Nissan vehicle owners and two sources of demographic data. With e.piphany's recent acquisition of RightPoint, Nissan hopes to add real-time communications with customers to increase cross-selling opportunities.

Owens & Minor (O&M). A $3 billion distributor of medical and surgical supplies, the company has turned its internal supply chain management data warehouse into a business for use by its trading partners, both upstream and downstream. Product suppliers and customers pay a basic subscription fee, plus the marginal pass-through costs of the end user software licenses, to get standard reports over the web showing purchasing contract compliance and buyer patterns for healthcare systems (who buy medical supplies through O&M), and service level reports for suppliers (who manufacture the products distributed by O&M).

The system features Business Objects WebIntelligence reports, backed with an Oracle 8 database that is loaded using Informatica's PowerMart, on a 6-way HP 'N' series UNIX server. The system has been gradually phased in over the past nine months and currently serves four of the top 15 suppliers and over 60 healthcare institutions. Since sales ramped up in the fall, O&M has signed up an average of 3 - 4 new customers per week. O&M's goal is to use the data warehouse to differentiate its services, with the fees intended only to recover costs.

Vlasic. A $1.5 billion food company, also recently spun off (from Campbells Soup, 18 months ago), the organization had a year to migrate off the parent firm's systems. That included the ERP system (AS/400-based BPCS). The data warehouse was initially used as a buffer system, which received extracts of master files from Campbells, staging it for loading into BPCS (PeopleSoft is used for HR). As part of an effort to improve marketing, the data warehouse was also used to generate sales reports. They use Brio as the web-based reporting tool, an Oracle 8 database, DataMirror for extracting data from BPCS, and Informatica's PowerMart to transform the data.

The system has proven popular with field sales. When the data warehouse was recently taken down for a scheduled upgrade, the IT group drew complaints; they interpreted that as a compliment.

PROJECT DRIVERS

In our earlier survey, we found that most projects were justified for broad strategic goals, such as organizational reengineering, customer service improvement, regulatory compliance. We also found some drivers were purely tactical, aimed at improving the speed or quality of decisonmaking, or having a more flexible alternative to standard hard copy 'greenbar' printouts.

The drivers remained pretty similar on this go-round, with typical examples including:

Consolidating financial reporting from multiple systems;

Tracking customer buying preferences;

Producing intelligible sales reports allowing individuals to track their performance against 30- or 90-day goals; and

Analyzing supply chain performance.

In one case, the goal was to turn an internal data warehouse application (on supply chain management) into a marketable service to business partners, as part of a broader effort to build a service business. 'The beauty of this was that we already had the database architecture in place,' said Don Stoller, data warehouse project leader for Owens & Minor, which built the system. As noted above, O&M will charge for use of the system, but the charges are mostly meant to defray costs. 'We saw this as more as a sales generator than a revenue generator,' noted Stoller.

In other cases, necessity became the mother of invention, as corporate restructurings drove three organizations in our sample to build data warehouses because that could be completed faster than full transaction system migrations. That proved the case at:

Dade Behring, where the organization endured two restructurings in five years;

Chicken of the Sea, which had been sold to its current owner two years ago; and

Vlasic Foods, which was spun off in 1998.

However, these newly-reconstituted organizations weren't necessarily just looking for quick and dirty reports. For instance, at Vlasic, the data warehouse was a major cornerstone to rebuild a marketing strategy that had been neglected when the company was under former ownership. Although most projects were motivated by a desire to improve on old, rigid mainframe reporting systems, in some cases, the old systems were just fine, or the user base was sophisticated enough to truly exercise them. At Dade, a significant number of non-IT staff were submitting SQL routines to get reports from the legacy system over the pervious decade. Hallmark had an adequate internally developed system before Microstrategy was installed. So why did these organizations decide to fix something that wasn't broke? At Dade, it was the necessity of building new systems after the company was spun off. At Hallmark, it was because the Microstrategy system offered superior functionality, and offered the side benefit of getting the IT group out of the software maintenance business, according to Tony Marshall, Hallmark decision support specialist.

Who drove the decision among our study sample? It was split almost down the middle between IT and business groups. For the companies spun off, IT had to take much of the initiative; at Vlasic, for example, the IT group was increased from three to 40 people, as it had to transition its transaction systems and build the new data warehouse. At Boeing Employees Credit Union, IT took a major part of the initiative because the systems crossed so many line businesses; a similar case was true at Kaiser Permanente.

But we also came across some grassroots efforts. At Chicken of the Sea, the CFO, a 20-year company veteran, was the champion of the new financial reporting system. At Imperial, the reporting system was also sponsored by the accounting and finance group. At Nissan North America, marketing carried the torch. 'We had to take control over the way we communicated with customers,' said Nissan's Ross.

OPERATION

Staffing. The most successful projects always included end user/business analysts playing central roles in implementation and maintenance, in addition to developers and, for more complex projects involving large or multiple databases, DBAs. That finding shouldn't be startling, because it's the same formula behind any successful IT project: a combination of business and systems expertise. 'You need good coordination between the DBAs and application developers,' said Hallmark's Marshall. Having both work side by side can help considerably when tuning such large systems; they can decide whether it's best to change the application or database table structure to get the right performance.

For instance, at Dade Behring, the ratio of IT technical and business analysts was usually 50/50; in fact, the project currently needs to recruit an additional business analyst. Chicken of the Sea used a developer who was familiar with the PRMS file structure and the legacy Query 400 reporting system. At the other end of the scale, the Imperial Tobacco project was driven by power users with finance backgrounds. The project leader, coming from the finance organization, took a bit of a career change in learning the Cognos tools, including developing the extraction and transformation logic, configuring the reports, and designing the OLAP cubes themselves.

Once the project goes live, do these projects retain staffing? In our earlier research, we found some projects to have sizable dedicated staffing, while in others, there was a part-time effort by one or two business analysts that gave as little as a few hours of monthly maintenance.

We found on average more commitment to data warehouse support this time around, but not without speed bumps. For instance, Dada Behring currently has four people to maintain the data warehouse, and is actively trying to recruit a fifth in what is a very tight job market.

Obviously, for large, well-entrenched systems, such as Hallmark's, a dedicated staff (in this case, seven people, including DBAs and developers) was already in place. At Owens & Minor, the productization and marketing of the data warehouse has provided a bottom line justification for the five-person staff, comprised of a DBA and several developers/business analysts. Given the competitive importance of both firm's efforts, and the fact that these projects have been established for several years, both had formal end user training programs. Hallmark offered two half-day courses, while O&M has a full day course for internal staff, and a half-day course for external customers.

Project Timeframes. We found a few 90-day wonders, but the median timeframe for delivery of most projects we studied was about 9-12 months for the first deliverable; after that, deliverables were usually planned for 90-day delivery. Boeing credit union's first phase was delivered in 9 weeks, with additions planned every 3 months after that. 'I want to build excitement, I want to keep rolling out new releases,' said Butch Leonardson, vice president of IT. He added that the rapid startup was made possible, not only by intensive training of the IT staff on Sagent tools, but liberal use of Sagent consultants for mentoring and jumpstarts.

Another exception was Hallmark, Although a huge system, Hallmark took only 6 months-but it had over 5 years of operating experience and a mature decision support system architecture when it began implementing Microstrategy. A recent project to add HTML web reporting took about 3 months.

Others took longer, for reasons ranging from limited staff commitment, resources stretched thin by concurrent IT projects, or the sheer complexity of systems begun from scratch. Imperial Tobacco's data mart was built by three finance organization staff members, who each contributed an average of 40% of their time over an 8-month period. At Owens & Minor, the rollout took 18 months because of the complexity of getting external users up to speed, and identifying and designing features of use to them. Meanwhile, at Vlasic, the data warehouse was being built at the same time that a new IT staff and data center was being built, and the manufacturing system (BPCS) was being migrated. Whenever emergencies happened with BPCS, the data warehouse project was temporarily place don the back burner. Miraculously, phase one of the data warehouse went live n only 6 months.

COSTS

Research Sample. Admittedly, small projects were under-represented in our sample; only one project came in below $100,000 (most were $250,000 up to $5 million).

In our previous report, smaller projects were more prominent. We believe that the fact that the project range was larger on this round was due to a combination of a freak of sampling and the fact that, as a maturing technology, more companies are investing more resources into data warehousing. Alternatively, many data warehousing projects are likely slipping under the radar screen, thanks to the bundling of OLAP services in Microsoft SQL Server 7, which allows projects to be developed within Microsoft Office and BackOffice-products that most IT organizations already have.

On the low end, Chicken of the Sea's data warehousing effort was fairly inexpensive; at around $10,000 (or less), the cost of Seagate Info was a fraction of more specialized competing tools that were considered, such as Silvon's Salestracker. The scope of the installation and the small user population (a couple dozen) allowed a workgroup-level solution which didn't require high-end functionality such as query management, security, load balancing, etc. For J.R. Simplot, which has ramped up its data warehouse slowly over the years, average annual costs (capital and staffing) have hovered around $250,000.

Other projects are costing well into the millions.

Boeing Employees Credit Union: A $750,000 investment was adequate for phase 1, which included building the basic data warehousing infrastructure, licensing the Sagent software, and getting liberal doses of Sagent consulting and team training to get the project off to a fast start. Thanks to the front end investment in training and consulting, and a well-established back end transaction system infrastructure, upcoming project phases should go on line at far less expense. Although phase one only served a handful of users (making the $750,000 figure look exorbitant), it was essentially a proof-of-concept; upcoming phases are designed to take the system into prime time, serving hundreds of users.

Owens & Minor: The startup costs for the Extranet data warehouse application hovered around $2 million; of that, staffing accounted for about 30 of the cost. They have spent an additional $400,000 to cover the forward costs of additional software licenses to accommodate end user growth over the next 18 months. By comparison, with the system in production, staffing costs have leveled at a $500,000 annual level (burdened).

Nissan: The cost structure was about $50,000 for hardware, $500,000 for software, and $750,000/year for the syndicated data that will become much of the lifeblood of the system. It takes about 2.5 FTEs to maintain the system; Nissan is currently looking for a database developer.

Vlasic: About $3 - $4 million for implementation of the basic data warehouse infrastructure and application. That broke down to about $170,000 for three dual-processor Pentium servers, $200,000 for software, with rest accounted for by internal staff effort. The large amount invested in implementation also reflected the need to build IT infrastructure from scratch, not to mention coordinating data warehouse and transaction system implementation side by side.

Dade Behring: They spent about $280,000 for Digital Alpha servers to run Oracle. However, many Oracle licenses were budgeted under other system projects (they would have been bought even without the data warehouse's existence). Overall costs of the data warehouse was about $2 million, including desktop software of about $1000/user.

ISSUES

Underlying the notion that every system is unique, we found that each project tended to have its own unique problems. While not statistically significant, together they provided a representative sampling of the types of challenges that data warehousing projects currently face. We discuss some of the most-mentioned issues below.

Deciphering Legacy Systems. Not surprisingly, this was especially challenging or companies that were bought, sold, or merged find themselves. They often inherit poorly documented foreign systems from the previous ownership. At Chicken of the Sea, navigating the data structures of the PRMS application inherited down the years was a greater hurdle tan implanting the new reporting system. At Dade Behring, it was dealing with the R/2 and R/3 systems of its newly acquired European operation, while at Vlasic, it was managing the transition of BPCS, plus importing master files from an aging system of its former parent.

Even where company ownership was stable, technology flux remained a formidable challenge. When Kaiser Permanente migrated from its managed care system from a homegrown application to a package, many data warehouse tables and reports were impacted. The primary cause was the decision to implement the new package as close to plain vanilla as possible. With the fields and tables structures in the package remaining largely unchanged, the data warehouse had to be modified to adapt. According to Jerry Nadjowksi, outgoing data warehouse project manager, up to one in four change management decisions regarding the new transaction system impacted the data warehouse.

Staffing. This problem is obviously not limited to data warehousing. In previous reports, we've noted the shortages of such key positions as DBAs (see Computer Finance, November 1997). We found several projects having to recruit new staffing. At Dade Behring, there was need for an additional business analyst. At Kaiser Permanente, there was the challenge of staff retention; once a business analyst grew proficient on the Cognos reporting and OLAP tools, they were often prime fodder for headhunters. That reminded us of the problems organizations faced during their SAP projects, as their newly trained ABAP 4 programmers got poached away by the Big Six (now Five).

Acceptance. The early 'Build It and They Will Come' philosophies fell victim to the ease of use problem. What we've learned is that a frequent problem of data warehousing is not offering too little-but too much.

For instance, Windows-based client/server applications, while easier to navigate than cryptic character terminal interfaces, can easily overdo the functionality. A well-known dilemma that virtually any user of Microsoft Office can relate to, several projects in our research sample reported that users were bombarded with interfaces that were too rich and complex.

This proved a key hurdle at Kaiser, where usage of the data warehouse varied from department to department, depending on the level of department manager buy-in and commitment to training. 'They had to crave the information,' in order to ensure that end users got the necessary training, said Najdowski. Of the initial target base of 150 users, Najdowski estimates that about 70% of the casual users were overwhelmed by the system. 'We didn't initially understand the needs of casual users,' he conceded.

Simplot learned similar lessons. 'You shouldn't overestimate the ability of users to pick up the technology,' said Bill Friend, vice president of supply chain management and information systems. 'It's relatively easy technology for those who are computer literate, but people who have trouble with email will have a hard me with this,' he said, adding, 'You have to shape expectations accordingly.'

In both cases, the problem may have been trying to do too much with technology before the right technology came along. In each case, having simplified web browser front ends and limited choices (e.g., a heavily constrained managed reporting environment with menus offering choices of a small, easy to identify group of reports) provide the solution. That was underscored by Owens & Minor's decision not to deploy the data warehouse externally until a suitable web front end was developed.

Technology. As mentioned previously, ease of use has been a major barrier. At Simplot, early data warehouse iterations using complex tools such as Sybase IQ and Platinum's Forest & Trees proved the wrong match. Consequently, Simplot tended to approach each data warehouse project tactically, without a long range architecture. 'We couldn't have done a planned investment back then,' said Friend, adding, 'If we did a long-range plan based on IQ or Forest & Trees, we'd be lost in America today.' He adds that, with the Brio reporting environment, the server architecture and the ease of use on the front end make him far more confident about planning ahead today. Friend notes that Brio has become the de facto standard reporting system for new projects.

Distributing data is another issue, where mobile staff are concerned. That's not a new issue. In our study of two years ago, we noted one food company that had its sales people use dial-up access once each week to download data via FTP. In our current sample, Dade Behring was typically emailing OLAP cubes in zipped (compressed) form. But even there, the solution is bulging at the seams, with many individual cubes sizes approaching 25 MB, compressed. That's too much for email, which via modem could easily take several hours to download, assuming there was no break in the connection. Dade is looking at web access, and ways to slice the files even smaller.

Another technology issue that is extremely relevant to ERP users is whether to go with the ERP vendor's own BI tools or go the third party approach. However, as we mentioned in our last report, offerings here are still immature. We didn't query any SAP BW (Business Information Warehouse) users in this report. But one of our respondents did actively consider BW, instead opting for Cognos's SAP Accelerator, which embeds data extraction and transformation technology from Acta Software because SAP's functionality was still incomplete.

RESULTS

The easiest results to document are the gems of identifying under-tapped markets or new sales opportunities. For that reason, it's not surprising that sales has been the most popular subject area for data warehousing. Our sample certainly supports this perception.

Hallmark can point to its detailed sales analyses as being responsible for showing that its new 'Warm Wishes' 99-cent greeting card product line would not cannibalize existing sales. Hallmark's analyses involve highly complex analyses; it tracks POS data, plus 21 separate product characteristics of each card. And, the company offers from 20,000 - 40,000 different cards at any one time.

Nissan has been able to use its data warehouse to finally pinpoint the buying habits of individual customers-filling a critical information gap when it came to identifying which customers are likely to be the most loyal. 'We've used the power of the tool to mine our databases and learn about our customer relationships,' said Ross, adding, 'The alternative would have been hiring statisticians to do modeling.' An early finding: their best customers were not part of the young male, moderate income demographic that most of their ads traditionally targeted. The stakes are huge; a 1% gain in customer retention rates could translates to tends of millions of dollars of added sales annually.

Vlasic's results are more anecdotal-but still real. The new system offers more data-90 day coverage instead of 30. The format makes the numbers easier to track and manipulate. It saves the costs of maintaining a call center for the 60 independent food brokers that used it, saving about $500,000 annually. But the real value comes in the form of new sales. They can point to at least one instance where a broker, armed with the numbers, was able to generate a client presentation while on the plane which provided such complete sales forecasts that the broker won an order that was 5 - 10% bigger. That translated to several million dollars of added sales, which easily offset the cost of the data warehouse.

Owens & Minor. The new data warehouse Extranet creates mindshare for superior service among its trading partners. Although the goal is to serve 1000 customers and 100 suppliers next year, bringing in revenues of about $2 million, the purpose is not to become a profit center, but to stimulate sales. That would come from customers ordering more of their supplies through O&M, suppliers making O&M their prime distribution channel, and from the sale of supply chain management consulting services.

EXPERIENCE COUNTS

Our sample, for the most part, consisted of mid-large projects built using best of breed tools. Absent from our research were entry-level efforts, using VB developers to write small data mart applications using Microsoft SQL Server 7's bundled OLAP services. Nor did we speak with users of ERP vendor data warehouses-most notably, SAP's BW (the most established offering in a very young field). We'll examine these projects in a future report.

Nonetheless, our sample was broad-based enough to identify trends in subject focus, implementation issues, and end user acceptance. Sure, we still noticed a preponderance of sales/marketing projects, but we also saw an upsurge in finance and operations-focused data marts. We found that end user training was a sensitive issue, and that, for older projects, the lack of web access was a barrier to reaching casual users. And, we also found that data warehousing projects were just as vulnerable as any IT project to staff turnover.

As for analytic applications, that market is just getting off the ground, and there is relatively scant user experience on whether these new solution-oriented data warehouses will provide faster and better returns on their investments. Similarly, there is little or no experience to judge whether outsourcing data warehouses to ASPs is feasible, given the fast changing nature of decision support. These are issues that we'll track closely over the next year.

In a follow-up report, we will compare our findings with the results of a recent Gartner Group study based on responses from over 100 users.

© 2000 ComputerWire Inc