Archives of the TeradataForum
Message Posted: Sun, 24 Jun 2001 @ 04:53:52 GMT
| Subj: || || Re: Why did Data Warehouse Project Fail? |
| From: || || Kent Clay |
Got to put in my buck-fifty...I see two major reasons why warehouse projects fail. One, Warehouses/relational environments are EXPENSIVE
projects if done correctly. Following the old "business" or "strategic" planning methods which, I feel, opened the door for the relational
movement, the All-time, A-Number-1, Whizbang requirement was support of upper management. Not the VP level, but executive VP level and
above. You can't freeze development for 1 - 2 years while a complete data study is done, nor spend the kind of money it takes to complete
this part of the project, without this type of support. Even those projects that have this support initially may lose it in the middle
because the sponsor didn't fully grasp what the real costs were going to be. I don't feel it's possible to have a successful warehouse
without a solid relational model in place supporting the operational data requirements of a business. Secondly, many relational projects
yield little more than an online "reporting" system similar to the VSAM/QSAM/Tape systems they replaced. Data modeling is not an easy
concept to grasp when you've done things the old way for so many years. This is along the same lines as programmers who tried to convert to
object-oriented coding. Studies show that it's easier to train a non-programmer in object-oriented than trying to un-train/re-train a
seasoned programmer. Data modeling is the same. Many companies jump into relational projects with their current staff. They may provide
some training initially but for the most part the project is doomed. It's just not going to happen using first-timers. The bad part about
it is that bringing in "professional" help doesn't ensure success either. There are a lot of people who claim to be whizbang modelers. For
the most part...they're not. How many people remember structured coding? Do you remember all of the definitions presented? Many thought
if they wrote a program with no GOTOs, it was structured. Ahhh, but we know better, don't we, gang?? If you have the opportunity to work
with a true modeler, it's like the clouds breaking up and the sun shining through...much the same as programmers I've spoken to about when
they first saw the object-oriented light.
I hope I haven't offended anyone, particularly those who feel they've designed the ultimate relational environment. I hope they have.
After all of these years, I'd still like a chance to again work with the person who "opened" my eyes.