November 8, 2004
(Computerworld)
Analytic technology is on the rise. Business intelligence software expenditures are a growing share of the IT budget. Data collection and storage costs are increasingly driven by analytic uses. DBMS and enterprise application vendors are focusing their product differentiation efforts largely on analytics.
Few organizations, however, are doing a great job of planning for the blossoming of analytic technology. True, many enterprises have a fairly careful design for the underlying data warehousing. But in most shops, departmental BI applications and analytic applications are installed almost willy-nilly.
It's time to take a more serious approach to analytic IT strategic planning, and not just because analytics represent a bigger part of your budget. Analytic technology simply has more parts than it used to, and many more points of integration, both within the analytic sphere and reaching out to transactional systems as well.
Let's examine five areas where analytic technology integration is of particular interest. In future columns, I'll talk about how to tie these and other considerations into a practical analytic IT road map.
Integrated monitoring, evaluation and information delivery. BI technology has historically included a mishmash of information delivery and analysis tools -- ad hoc query, ad hoc reporting, enterprise reporting, multidimensional analysis, graphical data visualization and so on. These are getting streamlined and integrated in a new generation of technology, and it's about time.
Over time, classic BI technologies are becoming less important anyway. The user's central monitoring tool will be the portal or dashboard. That format first shows which metrics are outside expected ranges and only afterward (if ever) leads the user to the precise numbers in a report. This is often an improvement over traditional reporting-centric systems, which may spit out a lot of data and then send the user on a "Where's Waldo?" search for anomalies. In cases where time really is money, alerts about the anomalies can be sent straight to a pager or other mobile device.
Monitoring, evaluation and transactional apps. BI technology has historically been read-only, running against a copy of the transactional database. So integrating it with transactional systems might seem technologically unnatural. But let's take a business process view. When a manager notices or is alerted to an anomaly in a metric -- then what? The "what" should often be a process leading to an action, perhaps in production or purchasing, but potentially in almost any area of the enterprise.
A whole new generation of hybrid analytic/transactional applications is emerging to support these new processes. You can wait to get them in packaged form, or you can roll your own, perhaps using some process specification tools. But either way, the processes -- and hence the applications -- are likely to be very important to you.
In-line analytics. A classic job for analytic technology is to figure out exactly what to offer to which customer, so as to make the relationship as profitable as it can possibly be. In certain environments, such as call centers at cell phone service providers, doing this analysis in real time is hugely important. So analytic tools -- often statistical ones -- need to drive transactional systems in an orderly way. Meanwhile, certain consumer marketing applications are trying to regularize testing and statistical analysis as part of, say, a transactional direct-mailing business process.
Planning and everything else. Almost every organization has a horrific budgeting and planning process. But modern enterprise planning technology has somewhat streamlined the process in thousands of installations. Even so, most enterprises' forecasts are still incredibly naive and poorly supported. As planning technology evolves, transactional applications, monitoring/evaluation, planning itself and sometimes even statistical analysis should be able to combine to create better and more timely forecasts, and more useful project plans.
Integrated analytic data management. There are some hard-core server technology issues to consider, too. Integrating enterprise reporting, ad hoc queries and various kinds of analysis into a single server can be a demanding task that bears close evaluation when you select analytics vendors. But the server-side issues are even broader than that. DBMS vendors are doing important work in data aggregation. BI vendors are doing their own important work to make the DBMS features as unnecessary as possible. Data caching integrates into application servers in interesting ways, and some prominent BI products include their own application servers. And vendors of specialized MOLAP (multidimensional online analytical processing) database servers are ever more frantically trying to find important uses for them as their core benefits are usurped by advances in relational DBMS technology.
That's just a sampling of the tough issues involved in setting your analytic technology plans. Next time, I'll talk about how best to deal with these and other analytic challenges.
Curt A. Monash is a consultant in Acton, Mass. You can reach him at curtmonash@monash.com.