Data mining is an intrusive activity and hence requires fool-proof planning. So in order to ensure that the user acceptance is easily dealt, you should start the implementation only after the following basic data mining preparation is in place:
Design the data matrix: First and foremost, get your data matrix right. You have to precisely know what specific parameters you are looking at. Remember, the standard template that has worked for a peer in telecom will not suit your manufacturing business. The parameters have to be drilled down to every possible detail. It has to differentiate anywhere and everywhere. This is critical for your data mining basics checklist.
Get the right schema: The second crucial aspect for data mining basics is that you have to get your schemas right, otherwise you won’t be able to mine what you are exactly looking at. You have to have your champion who can define how the data is interpreted at the business function’s grass-root level.
The business head is basically a champion of communicators. For instance, in a larger format retail store, only such a champion can provide you with usable data mining insights. Even the color of a refrigerator is a sensitive selection criterion for the customer. A specific color might be extremely popular in Bengaluru, but not in Delhi. Hence the point to be noted is that the CIO cannot perform an isolated function at all when it comes to the basic data mining requirements. Data mining activities depend on such basic aspects, and are useful only if the user can draw inferences based on this existing knowledge.
Budget seed money: You need to provision ‘seed’ money in your budget for the entire roll-out, especially the one that deals with data mining. You have to keep seeding till the plant that you want blooms. And this will be purely based on the success of the experiment. You have to articulate this precise logic to your management. So as part of your data mining basics checklist, you should remember (and reinforce) that funding is a continuous process.
Spot the pain points: The CIO has to be completely involved with people who created the design. For instance, he has to work with logistics to merely understand the impact of movement, and how it affects data. One has to be judicious in selecting an area for the pilot project. For instance, it could be procurement, where a wider insight in data collection provides a better opportunity to understand the trends. You have to go to the end user and find out what the pain areas in his regular day-to- day operations are. Tools chosen should be the outcome of technology and the voice of the customer. So corrective analysis is a critical part of the data mining basics checklist.
Proof of concept: After meeting people and recognizing their pain area, your next data mining basics activity is to select the right mining tool. The proof of concept is absolutely essential and all the interactions will surely help him in doing it. It can be single iterative or multi-iterative, but nevertheless it has to be iterative. Data mining activity is futile without support mechanisms and maintainability.
About the author: Anwer Bagdadi is a veteran Technology head in the industry. In his previous assignment, he worked as the senior vice president and chief technology officer at CFC India Services Pvt Ltd.
(As told to Snigdha Karjatkar)
This was first published in September 2010