Categories Software

Continued from Data Warehouse Components

Time and money. In a 1996 study published by IDC ("A Study of the Financial Impact of Data Warehousing," IDC, 1996), the average cost of building a data warehouse was $2.2 million, with an average time of 2.3 years to break even. Ninety percent of the companies in the study achieved greater than 40 percent return on investment (ROI), and 50 percent achieved over 160 percent ROI. The average ROI over three years, cumulative, was about 400 percent, with a higher ROI for data marts. Clearly, building a data warehouse can be profitable, but it's not a quick hit. Your company should be aware of the amount of investment likely before any payback can be expected.

Space. Data warehouses require lots of disk space. When estimating how much storage space you'll need, don't look at just the current production systems. Remember that you'll be building a system of historical data. Most organizations will want to keep at least a year's worth of data, and if you want to do any kind of trend analysis, you will need several years. Also, reports and analysis usually need more than one index. Be generous in your estimates of disk space. Data warehouses measured in terabytes are not unusual, and some large organizations are even using petabytes. (A petabyte equals 1,024 terabytes.)

Consolidation. Combining data from multiple sources may reveal incompatibilities or problems with your OLTP systems. Consistency is especially important for data-mining applications, since most data-mining tools teach themselves how to analyze the data by looking at it. You may need either to do extensive data scrubbing or to fix the OLTP systems. Talk to users about their data needs, and decide whether the result justifies the cost.

Security. Security considerations for data warehouses are different from those for OLTP systems. For a data warehouse to pay for itself, lots of users have to be able to benefit from it, and therefore more users will need access to data than are traditionally authorized by OLTP security. According to experts, a "right to know" mind-set must prevail over the "need to know" philosophy, which would restrict access to data warehouses and minimize their effectiveness. If your organization can't make that cultural change, a data warehouse may be a wasted investment.

User-friendliness. Data warehouses have to be user-friendly. Since a data warehouse is not a production system, users don't have to use it. And the fastest, most reliable system won't be used if the user doesn't understand it or finds it cumbersome to work with.

Project planning. A data warehouse sometimes takes years to implement. Before embarking on a data-warehousing project, determine what your business objectives for it are, what the potential costs are versus the benefits, what resources you'll need, and how much organizational commitment you'll need. Without good planning and user buy-in, the project is not likely to be successful.

As more and more corporations come to appreciate that the information they gather each day is an asset, they will rely more and more on data warehousing. But while a data warehouse can provide managers with the means to ask questions of their data and get back meaningful answers, it can't automatically make a company more profitable. A good technology can't substitute for good management. Whether a data warehouse becomes a valuable strategic tool or an expensive white elephant depends as much on the organization using it as on the technology.

More: Software

Published as Enterprise Computing in the 3/9/99 issue of PC Magazine.

Related Links
• ZDNet Enterprise