Info-Tech analysts are frequently asked a seemingly simple question: “Should we deploy a data warehouse before our BI initiative or should we wait?” While the question is simple, the answer is not. Info-Tech's guidelines assist decision makers through this challenge.
So What is a Data Warehouse?
A data warehouse (DW) is a typical element of many reporting infrastructures. They are common in large organizations and are becoming increasingly prevalent in the mid-market. Info-Tech defines a data warehouse as a dedicated software and hardware platform for integrating enterprise data from multiple sources. It uses scheduled routines for pulling data from transactional systems and for cleansing data. The purpose of a data warehouse is to establish a comprehensive source of enterprise data.
A well-implemented DW is a great boon to many enterprises. It can greatly facilitate reporting and analysis. But a DW project can also be incredibly expensive in terms of both time and resources. Many enterprises have started a reporting initiative with a DW only to find themselves several years and many dollars into the project without a single report delivered. It's hard to attain a positive ROI when there is so much upfront I(nvestment) with no R(eturn).
As an alternative to a DW, enterprises can run multiple data marts that meet the needs of a specific business unit or functional unit. These data marts are generally much easier to set up and lead to rapid implementation times. Furthermore, this quick start approach enables IT leaders to make early returns that justify subsequent investments.
Ultimately, leaders must address a single question: “Do we need a data warehouse?” In some cases a collection of data marts may suffice. Info-Tech provides some guidance.
Recommendations
A dedicated DW environment is likely necessary in environments with the following data characteristics and requirements:
- More than 15 data marts. Many enterprises begin reporting and analysis projects by developing specific business-unit solutions. These solutions serve as a low investment approach. Each of these data marts, however, requires a certain degree of maintenance. Enterprises with more than 15 data marts will likely benefit from the maintenance advantages of a DW.
- More than a billion records. Modern data marts can effectively handle millions of records. Dealing with billions of records is considerably more difficult and requires the capabilities of a dedicated DW.
- Historical data more than five years old. Maintaining the large volumes of historical data required for reporting on transactional systems can impair their performance. A DW can integrate archived or historical data that is omitted from transactional data sources.
- Tuned transactional systems. Pulling reports from transactional systems can downgrade their performance. This scenario can be particularly acute for transactional systems that must run a large number of batch processes during business hours or highly tuned systems. In cases where transactional processing is prioritized, the analytics environment may be compromised due to lag time.
- Poor data quality. Building a DW requires considerable attention to data quality and the management of master data. Enterprises that have experienced considerable difficulty with data quality may benefit from the discipline required of a DW.
- Frequent upgrades to transaction databases. Changing the table structure of a production database can initiate a cascade of problems in reporting systems. Each data mart and report will have to be checked and tested. The routines that feed a DW will still require updating but the cascading problem will be averted.
- End users confused by cryptic field names. It can be difficult to empower end-users to create their own reports in a data mart-intensive environment. One issue is the cryptic nature of field names in production data sources. Users simply don't understand which fields they should use. A DW gives designers access to greater descriptive depth and can improve the success of self-service initiatives.
- Senior executive demand for “one version of the truth.” Data marts are very effective for supporting the decision making at a business unit or regional level. There is increasing demand among executives, however, for systems that will support complex decisions by making all enterprise data available. Ultimately, if the requirement is for all data then a DW is probably required.
Bottom Line
A data warehouse is a great asset for many reporting and analytics projects. But it's also a very expensive asset. Use Info-Tech's guidelines to determine if a data warehouse is really required.
Souce: www.infotech.com