One reason that the very topic of BI is so nebulous is that it encompasses so many layers and technologies. There are the technologies you see, like reports and dashboards, and those that hide under the surface: data integration processes, a data warehouse, OLAP cubes, data mining algorithms, etc. Understanding how those pieces fit together (let alone managing it all) can be challenging. For this reason, ReLuminous advocates the concept of the BI Portfolio to manage and communicate a master BI strategy. Jill Dyche, VP of Best Practices at SAS, originally authored this concept, describing it as a set business capabilities incrementally deployed to an organization over time. This approach to a BI strategy is compelling for a variety of reasons.
- It communicates to an organization the notion that BI initiatives are not just about data nor the BI tools, but involve the coupling of information with business capabilities deployed across an organization over time.
- It emphasizes that, while certain BI applications may address a particular business function, the overall BI function is an enterprise-level set of technologies; the data warehouse is a shared resource across the organization. This reinforces the reuse of common components over a variety of business applications.
- It reinforces the reality that, for BI to be effective in the organization over the long haul, it cannot be viewed simply as project with a completion date. Rather, it will evolve as the organization, its strategies, and its goals evolve.
The Enterprise Data Warehouse
Our approach to BI in the multifamily space was developed and refined over several years while building the BI portfolio at Archstone. At its foundation sits the enterprise data warehouse (EDW), built in strict accordance with the Kimball Lifecycle approach that applies three fundamental concepts:
- Focusing on the business
- Dimensionally structuring the data that's delivered to the business via ad hoc queries and reports
- Iteratively develop the overall environment in manageable chunks that address a particular business process at a time
When implemented with attention to detail, a data warehouse becomes the foundation for delivering business value to the organization in the form of standard reporting, ad hoc analysis, dashboards/scorecards, and predictive analytics. Furthermore, a thoughtfully designed EDW provides the flexibility to evolve with the business. When managed with the input of a BI competency center (a co-op of analysts across multiple departments focusing on the use of BI in the organization), the results are quite powerful and inspiring to behold.
Feeding the EDW: The ETL Process
ETL stands for Extract, Transform, and Load - the processes that extract data from source systems, transform it via myriad business rules into the more business-centric form of the data warehouse, and then load it into the appropriate destination tables. The EDW design and implementation stage of any iteration typically amounts to about 70% of the effort and risk. This figure has a high standard deviation, subject to source data availability, quality, and complexity.
Delivering Business Value: Business Intelligence Applications
The best-designed data warehouse in the world will not deliver a dollar of return if is not leveraged to deliver value to the business. That is where BI applications come in, and they can generally be bucketed into one of four categories:
- Standard reporting
- Ad hoc analysis and data exploration tools
- Dashboards & Scorecards
- Data Mining/ Predictive Analytics
A well-conceived BI portfolio will address all four categories of BI application in the quest to deliver value broadly and deeply to the organization. For example, analysts and power users will appreciate the freedom to explore data "at the speed of thought" using ad hoc analysis tools, while community and regional managers will find more value in intuitive operations dashboards and scorecards that feature pre-defined drill-down paths. Senior executives will enjoy the option of monitoring an executive dashboard, exception-based alerts, or simply conventional reporting (using a single version of the truth as a data source). Finally, no BI initiative is complete without some data mining or predictive analytics application to fully leverage the information in the data warehouse.