Individuals involved in the business intelligence (BI) space often suffer the same fate as the doctor's children, who always seem to be the ones with the runny noses, or the experienced architect who is forever behind with the DIY on his own house.
As BI specialists, we get so caught up in delivering on business processes to the core business that we often forget that the delivery of the BI in itself is a business process. We are, after all, in the business of BI and as such, are so consumed by it!
Let me back up a bit. BI competencies around the world specialise in delivering analytics and reports on the various business processes. In essence, business processes generate measures within a certain business context. The measurement is the 'how-much' question, and the context becomes the answer to the 'for-what' and 'by-what' questions. So revenue (measure) by region (dimension context break) for product category X (dimension context filter) is an analytic question.
Typically, the question becomes: What would we want to measure in the business's BI? If we start with the ETL sub-system, we already can delve quite deep into some very interesting questions. The first of these, affecting the batch-window of scheduled jobs, would be the elapsed time to run the job. Going hand-in-hand with this (and being measured on the same level of grain) is the number of rows processed. Of course this happened on a given date, and if a company has good meta data, it may also be able to add the names of the developer/designer of the code, as well as the version that is currently running and the underlying business process as more dimensional context to the measurement. So, instead of having a 'flat' and quite boring batch-log table, a company now has a very usable star schema, that can be queried using any good ad-hoc reporting tool, where businesses could even build on an OLAP cube.
Trend setters
As BI specialists, we get so caught up in delivering on business processes to the core business that we often forget that the delivery of the BI in itself is a business process.
Cor Winckler is a technical director for the PBT Group.
This then immediately allows users to undertake a trend analysis - asking questions such as: Are we processing more records now than a month ago? This, in turn, can, for example, help with capacity planning, which looks at the question of if we keep growing at the current rate, when will we need to buy new disks? We are now able to answer these questions as a matter of course, instead of having to do a whole extract and analysis each time the question is asked.
In addition, another area worth looking at and measuring is that of data quality. If a company, for example, has a fixed set of data validation screens, these screens will generate measurements of how many records fail the validation test, out of the total number. The measurements here include the total rows and failed rows. The measurement may also include dimensional context like the source system name, table name and column name, the data owner, business department, the business process and so on. Again, if this is tracked and stored neatly in a star schema, one can easily see whether data quality is improving or deteriorating over time.
One can also analyse and compare whether certain source systems or departments have a higher rate of improvement or deterioration over time than others.
With all of this in mind, once we have given an organisation the opportunity to be intelligent by embedding an intelligence process such as BI into the business, we might want to pay more attention to BI of the BI, which for pronunciation reasons, we like to call intelligent business intelligence.
* Cor Winckler is a technical director for the PBT Group.
Share