Aug 2, 2010

What OLAP to use with SAP BW

Just came out of a meeting where the Houston based customer needs to make a OLAP delivery decision. They have been using BW for the last 3 years and are now migrating to Business Objects OLAP.

Decision 1 whether to stay with BEx reporting or move it all to BO

My recommendations:
1. There are many reports that are best run via the BW.
   a. Where performance is bad the customer had a BW Accelerator installed.
   b. I disagree with SAP’s recommendations that there is no modeling required for BW Accelerator
       i. BW Accelerator is a very expensive appliance and modeling can decrease BW footprint anywhere
          from 20 to 60%. This means reduction of blades by 20 to 60%, or being able to put 20 – 60% more
          cubes into the same number of blades.
      ii. The first choice should be to model for performance. We demonstrated increasing some DSO reports
          performance from 517 seconds to under 20 seconds in under 2 days of automated modeling. This is
          with no BWA or any hardware changes.
2. There are many reports that run more efficiently in Business Objects
    a. Core strength is Ad-Hoc capabilities
    b. Second core strength is Remote query capabilities
    c. Selection is critical
   d. Use Xcelsius for standalone visualization
    e. Use WebI for remote query capabilities and ad-hoc reporting
    f. Use Crystal Reports for fixed pixel reporting and fixed format reporting
The key once again is that OLAP deployment must not be simply a technology decision but that of business needs with OLAP becoming a competitive service provider to meet business needs.


  1. Your word in god's ear. I thought I was alone with my thinking.

    The biggest tradoff of BW is that there is no central defintion of a key figure (the one...).

    The whole history of DW/OLAP/BI reflects the possibilities of the hardware's speed in a certain period in IT and it will in a close future become a lot more vital to model for speed.

    A memory based column store is ok, but there is much wrong about BW/BIs underlying technology - SAP R/3 is built for anything else but batch jobs that I would say - it is all about batch jobs;-) and to forget about.

    PSA, ODS ... wonderfull, but coming back to Bill Inmon - I think he never thought about having an OLTP technology as the solid underlying for a datawarehouse.

    BW compensates SAP's inability to have a reporting built into the OLTP system. This is why a BW has nothing to do with BI from it's very whole concept - it happend. BW was in its early days something that should have resided on the OLTP system... So we have on the BWs tonns of data that answer every question or none...

    BO follows a tradition that comes close to Cognos ... This approach has prooven over the years. The staging technology can change not the reporting environment.

  2. My First book is technology agnostic, thus the big trade-off is in the whole concept of BI as it is right now.
    Whether it is BW, or Oracle DW or Cognos or any other BI the current rate of failure is 45%. Unfortunately no one knows that that 45% means.
    Doe it mean 45% of all BI implementations fail and are thus shut down (We know this is not true); does it means that in most BI implementations only 55% of reports are usable (Now that might be true- but not substantiated).

    From where I stand I think the current age we are living in is the new Information Age. Information is a bankable commodity and dealers of information will soon hold the world. The real question is whether delivering information is simply technology, a platform (the selling pronciples of BI) or understanding the needs of the Information Consumers and then building a solution to fit the needs (The marketing approach to BI)

    Bill Inmon rightly assumed that data was the center of his universe and DW, but as the technology has matured data modeling only gets one more data and not necessarily more information.
    BW or BO, Oracle or Informatica the solution lies with the consumer and not in the technology. Give the reigns to the consumer and watch the BI world explode like never before.

  3. I agree with many of your comments above however i do feel you are both forgetting one key aspect. Many BI projects fail due to the companies ability to see the main cause of the problem which is usually the validity of the data. Until organizations start to ensure that the data is right the will never get a return on their BI solution.

    BI projects fail for may reasons and usually they can be put into 3 main categories:
    1) Bad underlying data - no ownership of this data from creation to exploration
    2) No business buy-in - Having said that Gartner and others have said for many years the importance of a business stakeholder so is this really true?
    3) Poor system integrators who for many years have seen BI as a reporting commodity only. In recent years since SAP and others spent billions on bring in other organizations have they changed their model - many are still playing catch up but try to spread the word that they are true industry experts......

    BI is a journey and it takes time, effort and energy to implement correctly, but going back to what i said at the start 'you must get the data right before the BI works'..... There is an old saying 'building a house upon the sand'!!