Project research Question & the way ahead.

Class,

Recall the 3 basic research questions for the project. Of these, the most important IMHO, is Q2.

Here's why. Q1 'depends' on Q2 and Q2a. Q1 requires you to pull info from known and citable secondary sources to estimate a simple demand model. Marrying primary data into this is do-able but not straightforward. In any case, Q1 cannot be entirely answered without supply side information.

Hence, my suggestion is this: Just do a basic demand estimation exercise for Q1 and leave it at that. If you have the demand model ready, the variables selected, the data cleaned out and are facing implementation issues, pls, pls feel welcome to get in touch with me.

"OK. But what is demand estimation exactly?"
The classic demand model says, sales are a function of price and the other 3+ Ps. Sales Units = f(price, product attributes, promotion, place, preferences etc). Data on sales, on prices, and basic product attributes is sufficient to estimate elementary demand models.

"OK, but at what product aggregation level should this demand estimation be? Should I estimate the model at the brand level, the brand-model level or something else?"

Well, one possibility is car size - as enumerated in Q10 of the questionnaire. Another is to simply put together the top selling 20 models from the top 8 manufacturers and regress them on price, brand dummies, and a few major-major attributes like engine capacity in liters, mileage, fuel-type etc (which ones to include here can be obtained from the survey importance scores). And so on.

Whatever you do, do please write down your model in simple Y=f(X)+e terms. Clearly list the Y and the Xs, their descriptives and their nature.

"What does it mean that Q2 is very important?"
Q2 allows you to play with and use the survey data collected very nicely. Look up segments in the data, identify factors that explain behavior and preferences, identify ways to reach these segments and ways to exploit the factors and so on.

So purely from an application POV, Q2 affords far more leeway and learning with data, methods and modeling.

"Why is the project so quantitative?"
It's not really so. A lot and I mean a lot of the insights that derive from data analysis are qualitative in nature. For example: "What do the factors mean? What do the clusters represent? What does the segment profile mean?" These are essentially qualitative insights.

And some qualitative insights could be from other sources as well, not necessarily from survey data.

"What are the main take-aways from all this Quant we saw in the last 5 lectures?"
Understanding the how and why of model-building is probably the #1 take-away from the quant side of the course.

Designing or building a model, as a solution to a business problem is precisely what you have all along been working towards in Quant MKTR (whether you realized it or not). You now, as a matter of fact, have all the skill sets, the tools and the perspective necessary to design, build and estimate models. The moment you reduce a business problem to a Y = f(X) form, essentially, you will have built a model to solve that problem. Neat, eh?

And the best part is it is *not* rocket science. It is a very do-able challenge. Implementing a model, once it has been designed, is very straightforward using modern computer programs.

Model-building is more of an art than a science - what is the construct being modeled, which variables to have, what Y and what Xs, what f(.) etc. Pls notice that what goes into the model is a lot of qualitative understanding and input! Nobody (and certainly not the computer) can tell you what goes into the model, in what form and to what end - that is your (artful) contribution alone.

"OK what to expect ahead in MKTR?"
For one, don't feel too overwhelmed by all the quant in the last 2 lectures. A lot came at you but I'm told by the AAs that "the students have successfully managed courses as tough and complex as this one."

Shall ensure lecture 10 is a little more easygoing and relaxed than lectures 8 and 9 (as much for my health as for yours). We can have a good wrap up of all we have done - essentially bring it together and make it a unified, coherent whole.

Two, will design the exam and put up a few sample question types on blackboard so that you at least have some idea of what kinda questions may come. Have received multiple feedback on this count.

Three, am considering calling for a special 1 hour SPSS session on practical tips and implementation alone. Am not terribly good at SPSS myself, so am not sure how useful that might be. Friday night probably. Let me know if there is interest in such a session.

Sudhir