is completely different from one division to the other. Defining what exactly is the "customer" becomes difficult. Is the customer a company or is the customer an individual? Or is the customer that piece of building you are insuring?
There are good reasons for the differences in definitions, and there's no easy solution to how you create a corporate-wide common set of data definitions. The answer to the question of how you deal with that problem is not a fun one but it's the simplest one anyone has ever found: you agree to disagree, and you build separate data marts for each of the divisions. Then you find the few things that are common, like payables or receivables -- organizations that you owe money to or collect money from -- and you make corporate-wide databases with those. You get an added bonus when you identify common suppliers: you can draw all of your purchasing together to get one of the most important and profitable benefits of a data warehouse, wh
ich is being able to show your vendors all you're buying from them, which lets you get better discount rates.
BYTE:
What's another common 'gotcha', and how can you overcome it?
Paller:
Another one is one of my favorites, which is when a pair or group of vendors say, "Our products will work well together." It's quite common to hear from a vendor that it has a strategic alliance with another vendor and that the products will work together. Well, many times, they don't. So, the only time you get a strategic alliance that is viable is when it's client-driven. When a customer says "I will buy these two products if and only if they work well together and if and only if you agree to make the interface between the products part of your product." But the customer has to be mighty big to persuade a company to do this. You should also make the vendors prove that their products work together, and guarantee in writing that they continue to work well together
for at least three years.
BYTE:
What are other caveats for customers who've already rolled out their data warehouses?
Paller:
The reality is as soon as they are worthwhile, people, very important people, become dependent on them. And so all of the things that you have to do for a big business system become important. You have to have security that's real, you need scheduling to automatically update it and verify the update, and you have to have backup and recovery. You have to have centralized management for distributed servers. The amount of money that people are spending on that part of data warehousing will grow probably to 50 percent of all investment of data warehousing. Most people don't budget for this, but they should.
For more information on the Data Warehousing Institute, send e-mail to
tdwi@aol.com
.
photo_link (154 Kbytes)
Lessons Learned in the Data Mine
