There are two things I hate in discussions about real-time business intelligence.
First, pundits cite great examples of the business need for
real-time BI, but then go overboard by assuming that every report and
analysis needs to be done using real-time data.
The reality is most analysis is done looking at specific timeframes
(daily, weekly or monthly), trending (YTD) or period over period
analysis. Up-to-the minute data would be discarded or create “noise” in
analysis. The cost both to load and then to filter out the irrelevant
real-time data for analysis is much greater than most enterprises are
willing or able to spend. And it just makes things too complex.
The second area that riles me is that people, even high powered
architects who should know better, oversimplify real-time BI. As I
mentioned when I discussed SOA,
too often real-time BI is seen as solely accessing data rather than
involving more complex data integration. Other than accessing very
limited data such as data related to an individual customer, much
reporting and analysis involves gathering and transforming data from
many locations. This requires data integration rather than just data