I have always felt the BW fall short in delivering the true power of analytics that it was meant to do. Yes, you do have some basic data mining algorithm available in BW7.3 <= under the APD tool and even with BW 7.4, it is still lacking some serious predictive algorithm power for any hardcode data junkie to take it seriously.


Regardless of version, BW always had some basic and frequently used functions which at times can be useful in helping to address requirement whiteout having to download huge amount of data into an excel sheet or run third party analytic tools to perform the same type of analysis. Trust that SAP has done a great job in ensuring the complex mathematical formulas behind these functions has been done correctly.



However, as any respectable BW consultant would know, standard functionality can only take you so far and there will be a time when you will be challenge to provide algorithms which are not available and it is never a good feeling to leave your business users in the dark with no possible outcome or solutions. It makes it worse when they start downloading enterprise data into excel and prove you wrong with basis excel add on. As with everything SAP, the good guys at Walldorf did provide us an additional option to write our own procedure if none of the above delivered function is what you are looking for. But hey, how many of us out there can easily translate this into SQL Scripts?


This article is not about the ongoing battle between you and your users because as everyone knows, it is about taking them on a “journey” – I personally find this a cliché but it does embody the coloration that needs to happen to reach the end goal. I rather like the quote from Sonny from the movie The Best Exotic Marigold Hotel: Everything will be all right in the end… if it’s not all right then it’s not yet the end.


If you are running your BW server on HANA, you have a hidden gem concealed underneath that investment that your organisation has made. Some believe that data is the new oil – European Consumer Commissioner Meglena Kuneva and to mine for this new resource, SAP has developed a nifty tool call the Application Function Modeller (AFM). It is an Eclipse based modelling tool and has been long made available since SP6. As SAP’s partner we continue to see improvements in the form of stability and additions of algorithms with each new release to help us better understand our data. In my opinion, customers are able to finally benefit from a true single source of information without having to run third party analytic tools that source data from BW by replicates Info Cubes structures into their environment.


The AFM tool to a certain degree does away with the need to perform SQL scripting but of cause I am not suggesting that you entirely blindside yourself from not being able to interpret basic SQL commands – it is after all a database server with some exception to the norm. It just means that you are able to refocus your energy to explore and tweak the minefield of information that is available by using the correct algorithm and method to answer a specific business question. Imagine this, late Friday evening and you are being dragged into a meeting with folks from the marketing and operations department and you are ask to predict the outcome of a new product launch from the information that has been gathered. One approach to this is to use a decision tree to anticipate the market response so that your organisation can react appropriately. But sadly, from a BW implementation point of view, this does not materialise often enough and worse still, the team manning the day to day health and operation of the BW environment does not have any input into conversation such as these. What usually ends up happening is, you have an experience user from the marketing department demanding for a huge amount of data out from the reporting server and they start to perform data analysis on excel or whatever third party tool that is available. There is nothing wrong with that and I think they should own it – it is what they have been trained for, it is what makes them good at what they do. It is their bread and butter.

If you are not running your applications on HANA, fine, no contention here but if you are, I am certain the AFM tool will shine. Gone are the days of pesky external connections, constant nit-picking between IT and stakeholders when a simple structure has changed, away with long waiting period for what seems so trivial – data dump, if only they knew.  The list goes on and on and you can fill in with your own frustration here from whichever side of the fence you happen to be on.


Even though if you are not into the predictive space or unmoved by the hype that data mining is already upon us or is a space that you are not willing to jump into, just by having a graphic interface to answer business concern, puts the analytic power back into the users hand. This tool makes playing with data fun because it is so simple to use, given that you have gone through the standard documentation but apart from that, the performance is there, the tools works and the results are real.


I guess my parting thoughts on this matter is that BW has never been terrific at performing statistical calculation or running predictive algorithms without loss of sleep and the motivation that keeps you going is the inner drive of yours to do it at all cost with the help of an icy Red Bull by your lonely side. I would like to think that SAP has come to recognise this shortfall over the years and it became evident when SAP released their Predictive Analytics solution in 2011 and they continue to strengthen their market position with the acquisition of KXEN, who in their own space, is a market leader at what they do. With constant revision to the Predictive Analysis Library native to HANA, you can be positively confident that SAP will continue to make progress in this space. While BW excels in many other aspect of a data warehousing tool, mining for data is not its core strength and I find it comforting that they are other alternative offerings by SAP to address this gap.

http://www.cs.utah.edu/~jeffp/teaching/cs5955/L10-kmeans.pdf

http://www.cs.yale.edu/homes/el327/datamining2012aFiles/11_k_means_clustering.pdf

http://www.ibm.com/developerworks/data/library/techarticle/dm-1007predictiveanalyticssapspss/

https://help.sap.com/saphelp_nw04/helpdata/en/4a/eb293b31de281de10000000a114084/content.htm

To report this post you need to login first.

Be the first to leave a comment

You must be Logged on to comment or reply to a post.

Leave a Reply