creating new model

[nfldbunker]

hi

i have been put in charge of developiong a new cognos model for clinical efficieny. We have a ;list of facts that we need to develop in this model.

just wondering if anyone had best practises,docuemntation ideas on how to start with such a project

thanks

bunker

[blom0344]

Your question is a bit too general in nature to be able to give much input. do you need to model against a relational model or against an ideally modelled datamart?
Will there be a need to go against multifacts? Do you need to create a dimensional model on top of it (DMR)

There is quite a bit of documentation on best practice modelling, but the focus is mainly on the pretty ideal retail model.

The best starting point would be to give some more info from where you have to start with..

[nfldbunker]

Hi thanks for the feedback.

lets see...

Currently, we have a user who provides a bunch of facts to end users via pivot tables which is produced from a flat file taken from a coding/abstarating system 3M.

Now the data is available in SQL database and we want to create a clinical efficency model based on the data that exists in the 3M SQL database. This model would be able to produce the same type of reports which currently are produced plus other reports that could be built from this model either in report studio, query studio, etc.

Currently we are planning this development and looking @ cognos and creating a star schema on this model.

Any information would be greatly appreciated.

shawn 

[blom0344]

You indicate that the data is available in a SQL database. Do you mean it always was or is the data stored in a new way?
How did the user get the flat file? Can you relate the contents of the flatfile back to a set of SQL tables?
When using the flatfile to generate pivot tables which dimensions where used?
What is recorded as facts? Are these additive facts (like numbers/values) or semi-additive facts (percentages etc) Is there a time component at stake? (like the recording of transactions at a daily level) What is the granularity of the data? Can you identify dimensions? How are the dimensional values keyed?