Since college, it has been my practice to normalize data using 3rd normal form. But when I started working in the real world of programming, I've noticed the importance of denormalizing data by way of distributing information into tables which are actually related to each other i.e. date, companyid are both found or stamped in the parent table and child table; area, item classification of items with a maximum of 15 char width are directly stamped to the item table (no area id or classification id involved). That way, I could generate report specific for the child table without much relationship to establish. These method put me in questioned by upper management which I successfully explained though that it's an issue of "speed and functionality" that is at stake here and not on disk space. Recently, I was amazed by a sample database schema showing an invoice table with an stamped customer address, contact person, telephone, etc. What is this all about folks? Is this a good design or poor design?