Skip to Content

Visions for BI

The new year has started and everybody is telling his wishes and hopes for the future.

I’m a BI consultant for a while now and I was thinking, what could really help me to make my job easier. I agree completely with Vitaly in his great blog What should be the Next Big Thing for SAP in BW/BI?, user friendliness is certainly one. But is this all?

I’m dreaming of a transaction where you can really “model” within BI. Shouldn’t be software so easy that you e.g. drag and drop the whole dataflow from your datasource to your cubes or multiprovider? I could imagine that you have at your right all necessary elements like in a painting software. Take the “color” or component you need and drag it on the right. Start with a cube for example. Then connect your Cube with a DTP or transformation. Finally take a datasource and connect it to the transformation. Don’t care about the fields in too much detail. It’s done by the software. I know that’s a top-down approach. But isn’t it the way you start data modelling? You write down the requirements from your client. From here you start to build your cubes. In the next step you look if there is any business content available you can use. You install it and find out, “No, that’s not what I want!”. Now you start again with building your own dataflow. Helpfull would a software that recognizes what you want to do and supports you in your work.   

How can I imagine the future? You start again with the requirements written down in a document. Now you simply put this document in your BI system and say “create model”. Then the system reads through the document and creates the data modell. Cool, isn’t it? I admit the text has to be in sort of a “modeling” language. But that’s a future extension to BPEL! In the next step you add the source systems where your data comes from. Of course all master data is unified and no format errors occurs. You schedule the jobs for the next weeks and data is there where you want it to be. Sweet, sweet world….. 

I dream of an “Administrator Workbench” where you can easily access and drill-through to each component you want. No windows with very small tables, three rows and 5 columns and you would need at least the double. No way to resize or to rearrange it to your needs. I dream of a workbench that you can configure how you like it. Get rid of all the windows, menues and buttons you don’t need. Add the ones that you really need. I hope that Web 2.0 brings lot of improvements here.

Errors in data loading, sad of the same error every morning? I dream of a system that can be trained to remove errors. Datasource in your R/3 have changed and now all your requests are red… One of the next versions does the replication before it tries to load data!  

Ok, dreamed enough for the moment. Now I have to go back to my work and remove this silly replication error…..

You must be Logged on to comment or reply to a post.
  • …if we could change the text of a variable in the query, to suit what we are using it for. Sometimes we have to create the exact same variable again to just have a different name show up in the selection screen for the user. Having some kind of local property to change its text in the query would be a big help 🙂
  • Bring forth the business modelling and let the system figure out what the optimal technical solution is!

    Tedious data modeling and loading issues ruins the agility of BI because everything takes so long to do. BI Vendors need to realize that data modeling should be more business and data oriented, and not database design and administration oriented.

    – Why do we have load errors because the system can not load the data format into the table? I don’t care which format the system uses – if I’m loading data then the system should figures out how to define the table so if fits my data, not the other way around.

    – If I load lower case data into an upper case only field, why doesn’t the system automatically convert to upper case by default, it’s a no-brainer. I already decided in the infoobject definition that I only want upper case data in it…

    – Why can’t I just select a data flow and then the system automatically creates a process chain to load. It knows everything it needs to know, infoproviders, where to activate data, rollup aggregates, etc. Why do I have to do it all manually?

    – Why doesn’t the olap processer automatically create a mix of aggregates and cache so that it can identify parts of data needed in a query, e.g. I run a query for half of my business, and after that I run the same query for the entire data set. Why doesn’t the system identify that it has half the data in cache and only query the db for the remaining?

    –  I could go on…don’t even get me started with the entire setup of extractors that need R/3 customization when a new field is added, super!

    But if everything was that easy then we would all be out of a job – or would we. Maybe we would spend more time realizing the potential of BI and business benefits, not in the juming though hundreds of transactions in an endless troubleshooting cycle.