One data mart to go? - prepackaged data marts - includes related article on the success of suites - data warehousing directions: the next generation - Buyers Guide

Software Magazine, Feb, 1997 by Teri Robinson

They haven't exactly erected the Golden Arches of data yet, but database vendors have begun serving Happy Meals of sorts to would-be data warehouse architects. Like McDonald's, which built a business on packaging value, vendors are trying to satisfy customer demand for fast data -- data marts that can be set up in 60 to 90 days with an acceptable return on investment following shortly thereafter. These prepackaged solutions include everything from server software to development and access tool to professional consulting services.

Looking at the numbers, it's easy to understand why most companies are hot to get a data mart in place. Last year International Data Corp., a market research firm in Framingham, Mass., examined 62 companies with data warehouse applications in place and determined that the average return on investment (ROI) was 401%, with a mean payback period of 2.31 years. The vendors and buyers of quick-start data warehousing packages hope to extend that impressive ROI to the data mart.

But just because the vendors are dishing up quick fare doesn't mean that you should buy the first prepackaged solution you come across. Buyers should beware: Some data warehouse vendors haven't put together consistent meals for data-hungry customers, the quick start programs may not scale, and you will need to customize them for your operation.

First off, be aware that the so-called quick-start packages are meant to develop data marts, not full-blown data warehouses. It's impossible to shoehorn a complicated, large-scale data warehouse into a shrink-wrapped box. At their best, quick-start packages are sets of tools and services that do as promised: help IT departments deliver data marts fast at a much lower price than if companies purchased the individual parts separately. "The price just makes it more attractive," says Chris Lange, database administrator and data warehouse architect at National Semiconductor Corp., which is based in Santa Clara, Calif. NSC individually purchased the elements of its 40-Gbyte data mart from Sybase, Business Objects, and IBM; it plans to buy the PowerMart design and management toolset from Informatica. Lange had purchased on an a la carte basis roughly the same package that Sybase now offers in its QuickStart package for $125,000. While he declines to discuss costs, Lange indicates that he would have spent less if he had bought the bundle. At their worst, prepackaged data marts don't do what buyers expect. The tools may not work well together, the products themselves aren't always scalable, and they may require significant recoding. And they're definitely not for building big, complex systems.

According to Larry Bramlett, formerly a data architect at Pacific Bell, most package deals are insufficient for a company with complex data warehouse needs like PacBell. "Anytime I've gotten involved in precanned software, it doesn't work," says Bramlett, who recently left PacBell to form his own consulting company, Data Warehouse Solutions in San Ramon, Calif. "Most are built to fairly generic specifications and [you end up] recoding and re-engineering."

Pricing and packaging vary greatly. For example, for around $30,000 you can get Pilot Software's Decision Support Suite, which includes the Pilot Desktop, Designer, Analysis Library, Excel Add-In, Discovery Server, and Internet Publisher. Compare that to the $125,000 price tag attached to Sybase's offering: a five-user license on Sybase IQ; PowerMart from Informatica; development tools from Cognos; access and analysis tools from Business Objects and Brio Technology; plus professional services and training.

Newcomer Sagent Technology Inc. of Menlo Park Calif. is offering the Sagent Datamart Solution for $50,000. IBM's $22,950 deal, called Visual Warehouse Solution, consists of its Visual Warehouse server (which manages multiple databases with a single dictionary), DataGuide for managing look-ups of shared data, IBM Database Server, Lotus Approach for end-user access, plus utilities and ODBC drivers.

Not all companies, of course, are trying to rip a page out of the fast-food cookbook. Unlike the product packagers, data warehouse pioneer Red Brick in Los Gatos, Calif., is focusing on offering services. "We've resisted the temptation to do package deals," says Steve O'Brien, director of product marketing. "It's difficult to believe that every data warehouse will be built with the same three tools." Instead, Red Brick will help plan a tailored data warehouse that can be deployed rapidly.

All of the prepackaged solutions are expandable, of course. How much is the question. "Normally, they don't scale very well," says Stephen Cranford, a partner in the consulting services unit of KPMG Peat Marwick in Baltimore. This may be more a matter of intent than product limitations.

Scalability becomes an issue because a successful quick-start implementation that generates a rapid payback attracts other business units. The resulting clamor to expand the scope of the rapid implementation can stress the system, notes IBM's George Zagelow, program manager, data warehousing solutions, in San Jose, Calif. "A quick-start package isn't enough -- there must be a scalability plan."