One of my clients has just decided to transform his methodology for creating new applications. I could not be happier about the decision. The client will finally address the fact that detailed specifications cannot be completed until the physical data model is created. Now, that does not mean the "detail" team cannot start on those detailed specifications – especially for mapping, integration and conversion purposes. But the physical data model needs to be completed for those specs to be considered done.
How many of you have gotten so close to production, and found some data issue? I have! In my experience the closer you get to production, the more a change will cost in time and money. I have been burned by teams and methodologies that want to get an application put into production BEFORE ITS TIME! If the methodology called for data profiling the source's data and understanding any integration concerns UP FRONT – that would be so wonderful. But, not all methodologies require profiling.
I believe it is important to understand any data quality and integration issues or concerns up front, BEFORE THE MODEL CAN BE FINISHED. If profiling is included in the methodology, the findings could be included in the data model (or required documents) at the beginning and not added as an add-on later in the project where change costs so much.
So, I guess the data model can’t be done until the profiling is completed, and the details specifications should require both.
ADD profiling to your methodology - trust me!