Everyone talks about data warehouses when it comes to business intelligence (BI); but seldom about data marts. They mostly ignore the term ‘data mart’ even though it is what they are actually working with. Even yours truly is guilty of that more often than not. Alright, so what’s so ding-dong special about a data mart? And what’s this fuss that I’m trying to pull?
Well, a data mart is just the same as a data warehouse, except that it only focuses on a certain aspect of an organization. Say you are a manufacturing company with many departments and branches; you will create a data warehouse that revolves around providing BI to the sales department only. It will only possess data that is relevant to the sales department. This is a data mart. Similarly you may create other data marts for other departments of the business.
The point is, when you create data marts for the various aspects of your business, you need not waste time, and can get things finished faster and with a higher success rate and also get to use newer technologies for a better user experience. Too many a time data warehouse projects have failed because it takes a long time to gather requirements, build and test. This is because they focus on the entire business. And when they finally think that they have finished, a life time has passed: technologies have changed (competitors are seeing data in fancy 3D applications, while you stare at monochrome reports), the original requirements have evolved into something entirely different (One; because that’s what suits always make you do. Two; because people have left and others have joined the company), conflicts (this department does not want this while that department wants it that way – and you spend weeks trying to mediate and get them to settle on something both can be happy with) plus there’s more which I cannot remember right now. Ah yes, and politics. Politics within an organization always puts you on the backfoot and delays work. These are reasons why data marts work much better.
The downside to data marting is Integration; such as when the big boss wants to see data from the department A data mart and the department B data mart and do some sort of comparison. Well, then you are in a fix now, but not as big a fix you would have been if you were building a data warehouse to cover the entire organization’s data. This is when you as an organization have matured. You are using effective BI day to day, though isolated to each department. You can now look at steps to integrate your data marts, into one beautiful data warehouse. Until then, it’s data marts and data marting.