What Data Mining Can Do and Can't Do

Jim Novo has written a blog entry cautioning the marketers for the use of analytics and has rightly pointed few issues in data mining and analytics application in business. I am taking forward the discussion here. (The original discussion started here).

I agree with Jim on the issues he has raised and his concerns are genuine. But I would like to add more depth in the use of the data for analysis and modeling purposes, and also its use in actual business scenerio.

As Jim said, segmentation may be highly predictive but may not be useful. To make it useful is a challenge and we need to take up the challenge. Just saying it may not be useful is not the solution. This just undermines the very basic use of analytics in Marketing space.

I have found objective and non-objective segmentation useful in different cases. First one may give good business insight but may not be actionable while the second gives very natural segments and actionable but of very little business gain. I would say, a combination of both is the necessity of the hour. We analyst have to be able to give business related natural segments. The segments then have to be profiled well so that the business manager does not need to dig analytics jargon to implement the data driven strategy. And this is all possible (may be hard).

Regarding models, I have seen a combination of few models is very necessary in any business scene. While talking of banks, I would suggest Revenue Model, Risk Model, Response Model, Bad-debt Model etc to be used together. The importance can be given to the objective that is more important at a particular point of time. The business needs may be to acquire more customers or increase profitability; it can be to reduce attrition or to improve ROI etc. We analyst must be able to give well designed strategy to the marketers that is data driven and the less we expect from marketers the better. Simply building models or segmenting the population is not enough.

So, my say is analysts/consultants need to do more while the data base team need to collect more data (not less). Marketers need to explain the business goals to the analysts and should not shy in seeking their help for successful implementation. Afterall knowing customers is the key and data tells it better than expert opinion, all for lack of psycological bias.