Thursday, September 22, 2011

An Alum wrote in...

From the Class of 2011. Am disguising Identity for privacy reasons.

Dear Professor,

I was wondering that how can we apply concepts and tools used in MKTR in B2B settings.

I am currently working in a management consulting firm(*hidden*) which helps foreign companaies formulate and execute their India entry strategies. I am in the industrial verticals which mainly deals with clients in steel,chemicals, auto and other core industries.

For formulating the strategies, we do market research as well. I was wondering what all concepts and tools can be used in such a setting when data points are limited to under 30-40. for example if i have a client who is a industrial chemical manufacturer, his end-users, competitors, distributors, logistic partners are not more than 30 each in India. After having done the research and collected data in terms of size, volumes, revenues, prices, quality etc., what all tools can I use to analyse and present the findings to the client?

Looking forward to your help in this.
Thanks,
A

 
My response below:


Hi A,

Its always nice to hear from an old student and know that MKTR gyan from ISB classrooms is finding place in real-world practice.
Sure, an entire array of MKTR tools are equally applicable to B2B scenarios as to B2C ones. The only difference is that we use 'firm' instead of 'consumer' or 'respondent' as out units of analysis.

From the one example you have described, it does seem to me that you have firm profile data in a sector. Depending on what your objective is, such data can be used in various ways.

For instance, if you wanted to segment the market into groups of similar profiled companies, a simple cluster analysis algorithm (e.g. the humble k-means) could classify the firms into k groups. You could then profile these groups in terms of their dominant characteristics. Depending on which group your client falls in, you would know which firms are likely to be the closest competitors to your client. If the Q arises, how do I know which variables are most relevant or important to use in the cluster analysis example, you could consider (a) regressing sales over these variables to see which correlate most with sales, or (b) performing factor analysis on the variables you have to see if there exist correlated factors underlying the variables you have.

Going further downstream, a simple positioning map is not hard to do if you have two or more ratio-scaled dimensions identified.

And so on. It all depends on what the decision problem is and what data you have. Besides, you don't really have to worry about central limit theorem defined sample size limits unless you are doing t-tests or the like. In any case, newer methods have evolved (e.g. Bayesian analysis) that do not invoke the CLT and perform exact small sample inference.

Hope that helped.
Regards,
Sudhir



 

No comments:

Post a Comment

Constructive feedback appreciated. Please try to be civil, as far as feasible. Thanks.