[MUSIC] All right folks, welcome back. Well in the past video, you would've seen this reference made to problem formulation skills, as the core competency for market research. For research projects, in some sense. Wherever you're going now, is to basically build on what happened there. And in some sense, come out with a framework for problem formulation. Where we are going now is problem formulation. Now remember, data analytics, well analytics is about solving problems. And before you solve a problem, it does make sense to figure out what the problem is, in some sense, to formulate the problem. In fact, a problem well formulated is half the job done, right? Half the job of solutioneering it done. All right, so let's get there, let me start with actually a traditional business problem, right? Traditional as in old economy, 20th century. Take a look at this, musing of a decision maker country held for XYZ corporation, give it a read. Quick read, take a minute. Okay, so welcome back. What you saw there, is an old data pickle right? This happens in firms of all sizes and shapes, in different industries all over the place. Which brings up the question, so what exactly is the problem here? Can we formulate it, can we put it into a defined structure that would enable finding a solution, okay? First question, is reality orderly? The answer is no. And we have no right to expect it to be. Coming from the academic world, the ivory tower perspective, typically, you see textbooks and questions in textbooks having a neat boundary around them. In the real world, there is no such boundary. Two, the medical analogy. And this is an interesting one, right? Typically, when a patient goes to a doctor with a problem, what the patient actually show the doctor is not the problem itself but the symptoms. One is running a fever, the most common symptom probably the doctors would face, primary health caregivers would be fever. Similarly, in the business world the most common symptom that businesses probably face is falling sales. Remember, sales falling is not the problem itself it is the symptom. The problem lies somewhere below the cause which has to be in some sense diagnosed. Okay, which is where the miracle analogy is coming from. How do symptoms relate to causes in the example we just saw? In myriad ways. Okay, what is the symptom there? The symptom is basically the sales were falling. What are the probable possible causes in there? And, I can count at least three, maybe there are more. We will see how they relate to one another. How do you diagnose a problem or a cause? Right, so we're not talking about symptoms which are observable, how would you diagnose a problem based on the symptoms alone? If you're a doctor, what would you do? Someone comes to you with a fever. One of the first things you probably likely to do is check for chest congestion. Why is that? Basically, it means if chest is congested or maybe it's bacteria otherwise it's viral. So, what you're doing is, you taking some immediate steps in the here and now, okay? Which it can be done right away and trying to eliminate some causes so you're trying to narrow the field, right and that is an important way in some sense. Well, by which we would attempt to structure messy reality. What data might be needed for the problem? That crucially depends on what the problem is, so let's get there, right? In the ideal case, right, individual causes would be isolated. So, one cause has nothing to do with the other, they don't appear together, they're completely independent. And each case would raise clear isolated symptoms which would make it easy for me or these are the symptoms then this is the problem. Unfortunately, the real world doesn't quite work that way. If the deal case were true then, what happens is that the problem becomes what we call separable. They are independent. They can be treated as distinct separate box. If the problem is separable, each part could be treated and analyze separately. And if that would possible to do then we could reach our decision based on the evaluation of each part. However, there is a data reduction dimension here, I'll come to that again. However, in the real world, what are the odds that the parts of the problem are interconnected or not? If the problem becomes separable and breakable into smaller pieces, there are hopes that, well, if that is not going to happen, at least certain groups of parts are separable. We'll come to an example for all of this. Groups of interconnected problems would have coherent solution strategies. Again, these are hopes. There is no guarantee this will happen. We will see. And, how do you reach a decision based on the evaluation of each piece? Each problem group becomes a hypothesis, which you can collect data for and in some sense test true or false, yes or no, and that would aid decision making. Okay, so let's see this. Consider a list of possible causes in the example we just saw, right? One, the product line is obsolete, it's just not selling anymore, competition is doing better. Two, the customer connect is ineffective. This may or may not be related to problem line being obsolete. Three, product pricing is perhaps uncompetitive. And this again, may not be related with anything else that is going on. The fourth probably cause is the sales force compensation policy that went off, and so on, right? Let's stick with these three for now. So, you have messy reality and then, you have a decision problem. In between, by actually breaking it into this list of probable causes, we are trying to impose some preliminary structure onto this messy reality. When we do that, we arrive at what are called decision problems. Here is an example for a decision problem or a DP. If number one is a probable cost. Product line of obsolete, the DP would say, should new products be introduced? If number two is the probably cause, the DP would say, should advertising campaign be in some sense changed, all right? If number three is should product prices be changed and so on? However, it doesn't stop there, okay? That the problem formulation process continues and drills down to something much more specific, something that we call a research objective. Why, because the DP may not contain sufficient information that would map directly onto the analytic toolbox. Remember, end of the day, what we want to do is define the problem well enough, so that we can use that analytic tools to solve them. And messy reality confronts you that it's hard to find structure there at all, we are imposing the structure, breaking it into parts, trying to find a way whereby an analytic tools can help us mitigate the size of the problem, bring it to manageable dimensions and finally, solve it. That's where we are headed. To do that, we need one more level of refinement. Then, the decision problem we need something called an arrow or research objective. And, that is where we are headed next. [MUSIC]
