Features
Feature
4 Enterprise
BI Realities
Lessons learned about operational BI accelerate bottom-line growth.
by Edmond Segalen
Nothing is better than a successful hands-on deployment of a business intelligence (BI) application to check market status. In working on such projects built around data warehouses for business and government, I’ve discovered that each deployment, in some way, has been a microcosm of the overall enterprise BI market.
Granted, no single implementation can teach you everything about a market as dynamic as BI. But it can provide lessons that can make the next deployment more successful. Here are four truths gleaned from working in the field.
1. A New Front
Until recently, most BI deployments leveraged a data warehouse for strategic purposes—designed to deliver decision-support tools to top management in the form of reports for long-term planning. Today, most implementations involve front-line operational systems. This requires real-time analytics for operational business decision support.
Embracing operations changes the way you think about BI and data warehouses. Performance, of course, is essential. The most significant change, however, is that BI application developers must establish business rules and processes in the operations involved. If you miss a step in operational business processes, the application generates multiple exceptions. When too many exceptions occur, users will ignore the program, possibly to the detriment of operations, and, most certainly, the application will be deemed a failure. It is vital, therefore, to rigorously deconstruct every business rule and process, including internal and external governance mandates, up front.
In Europe, for example, telcos use BI to analyze call detail records (CDRs) to detect fraudulent use of services. However, by rigorously adhering to governance mandates in the application, police with warrants can use the same CDRs to track criminal activity.
2. Business Pressures Proliferate
The growth in operational BI is feeding a need for more rapid time-to-results on BI/data warehouse deployments. There is little time to waste when such work leads to greater efficiencies or new revenue streams in day-to-day operations.
Spending a bit of extra time up front for testing a proof of concept (POC) can save much more time when the application rolls out. One way to improve that is to test with real user data. Naturally, many users are reluctant to hand over true business data for a mere test. But don’t let that reluctance prevail.
For example, recently a user was hesitant to hand over sensitive customer data for a POC test so such details were masked using a packaged application. Because some business rules were data-specific, leveraging actual data shaved off almost two weeks from the POC phase of the project. Tuning the system for simulated data would have taken much more time. Plus, performance benchmarks in the completed POC more accurately matched real-world results.
3. Unstructured Insight
Unstructured data, such as e-mail, PowerPoint presentations and other documents, contains valuable business information that was once the exclusive domain of specialized enterprise content management (ECM) systems. But ECMs cannot deliver the insights BI can, so users are pushing for them to get smarter or for BI to embrace unstructured data. However, in major markets such as pharmaceuticals, finance and government, much of that unstructured data falls under regulatory guidelines and, thus, cannot be transformed to work in traditional, structured BI/data warehouse environments.
Companies need to sift through customer feedback found in the voluminous content of e-mail, social networks, such as Twitter, online forms, and other unstructured sources. They want to detect trends as early as possible that could affect the business. For instance, negative or positive commentary on a social network about a new product could affect inventory levels before the data appears in an enterprise resource planning system.
BI must evolve to meet the unstructured data challenge. Users should be able to leverage BI systems so that queries to a data warehouse can include the knowledge latent in unstructured data and combine it with information residing in structured data.
4. Exploiting Spatial Data
Increasingly, organizations demand to understand data about objects in the real world, such as roads, buildings and city boundaries. They want this geospatial information for analysis. Business users have discovered that spatial data yields actionable and profitable knowledge.
In a recent deployment, analysts at a large European transportation agency predicted revenue-level shifts based on the different fares consumers would pay depending on route changes for public transport. This kind of knowledge enables better budget planning based on necessary route changes or the alteration of planned routes.
Gleaning business knowledge from geospatial data will broaden to include logistics, retail, telecommunications, manufacturing and other realms where relationships among objects, places, and other dimensional and non-dimensional data meet.
The Bottom Line
Obviously, other BI/data warehouse application development truths exist. However, these four are all significant, and their influence is helping drive growth in BI deployments, which will lead to bottom-line improvements in business. So, while strategic BI deployments will continue to grow in number, operational, dollars-and-cents BI applications will accelerate at a faster pace.
Edmond Segalen is an architect in Business Information Management within Capgemini.