Point taken :) To use a metaphor, the concept of BI is similar to a "database in the sky" in which you plug a crystal globe that should give a glimpse into the future and answer all the questions you did or never asked.
The description is less far-fetched than one might think. I'll explain shortly why and what type of FOSS projects you can use to make the concept a reality. The term Intelligence is borrowed from the military and is used to refer to the process of gathering and evaluating of information. In the business context this can be interpreted as gathering information about self and the surrounding business environment to evaluate one's commercial and economic context. How is this achieved? By measuring several parameter values that define the business processes, called key performance indicators - KPIs - and the influencing factors that determine these values. When you are interested only in the values of the KPIs, canned analysis reports are enough. For this reporting systems like Jasper or BIRT or any visual statistical project should do the job. These reports tell you how the business is doing. However, when you want to know the factors weighing in these values, using canned reports in not enough. You need to be able to click on the value and see its composition and formula and perform such operations like drilling, decomposing, dicing and slicing (and others) in the available data. In all these operations the visual element is paramount. These are so called BI Portals that include dynamic analysis reports, dashboards, scorecards and other BI appliances. For this Pentaho is the best in my opinion, but there are other projects easier to configure and maintain like OpenDX. All these reports tell a business what it can do, when and where, to have the most effective impact in improving the activity and profitability. In analysing the history of an activity, like sales for example, you may want to know product associations, groupings, and other characteristics that can be foreseen by projecting patterns discovered in the activity's history. For this there are Data Mining tools like Weka - NZ made - packaged by Pentaho in its BI suite, RapidMiner or any other machine learning application with visual output. This is useful to show what worked best and what didn't, and how is it likely to perform in the future. These three elements, Reporting, Dynamic Analysis and Data Mining, make the crystal globe I mentioned in the beginning. What follows is the elements of the "database-in-the-sky". All these KPIs are highly aggregated values and as such have to be supported by a type of data repository that can handle multiple summing and other operations on sets (tuples in multidimensional terminology) fast. This is the job of OLAP databases, of which Mondrian is the only business-grade project in the open source space I know of, but to be fair, Palo is not far behind. OLAP databases are usually fed from Data Warehouses, which are de-normalized databases that store the raw data. Here either MySQL or Postgresql are good choices. Ultimately the data in DWs is loaded from one or more operational applications - applications that serve the operations of the business, like OfBiz or SugarCRM, hence the classification - through a program called ETL, short for Extract, Transform and Load. Here Talend is the benchmark for this class of projects which also includes Camel, Palo ETL and ETL Integrator, just to name a few. Pretty much that's it, in a very simplistic way. Mind you, to have BI quality analysis, you don't always need all these programs. Disclaimer: All the programs mentioned here run on Linux. Some run only on Linux. Cheers, Adrian On Thu, 2009-03-05 at 13:08 +1300, Nick Rout wrote: > On Thu, Mar 5, 2009 at 12:28 PM, Adrian Mageanu > <[email protected]> wrote: > > Hi, > > > > I can see that the enthusiasm is not too overwhelming for this subject, > > being only loosely related to Linux (David not included, we can talk > > about it off list) > > > > Nevertheless, in the absence of other topics, I can go ahead and do the > > talk, in which case I will need a confirmation for the booking. > > > > Otherwise I can quietly clear the scene to make room for other > > activities. > > > > In any case please let me know by tomorrow morning, on list preferable. > > > > Cheers, > > > > Adrian > > Perhaps if someone explained in plain language what this business > intelligence software is supposed to acheive, it might appear more > interesting. (Or not LOL)
