In the last video, we discussed, how our decision domains such as market versus product, dictates the analysis and the data structure we need for such analysis. In this part, we will see that even within a domain, there are various levels at which a manager may want to act or decide. Thus, our analysis and data structure may further be determined by that. For example, let's zoom into one of the domains we discussed in the last video, that is, product. Let's say we want to take some action and decision within the product domain. Even within the product domain, you will notice that there are various levels at which the manager may want to act. For example, at the highest level, the manager may want to decide which departments to put on promotion and which not to. At little lower level, they might want to tweak their promotions based on the category. Going one level further down, they might want to come up with a discounting scheme, based on each line item, such that each line item may receive different discount percentage. Let's say we want to take a decision at the line item level. Then, as discussed in the last video, our data must be such that each row represents a separate line item and the columns capture some characteristic of line-item. The characteristics could be price, cost, sales. Thus to reiterate, we must decide the domain of decision and then within the domain we must decide the level of analysis, like we focused on decision in the product domain and within that, we decided that we wanted to offer a distinct discount for each line item. To ensure that you get the concept well and at the risk of being repetitive, I will take example in another domain. Let say, we want to take a decision in the domain of time and not in the domain of product. That means, we want to customize discount percentage not based on product, but based on some unit of time. Now, within the domain of time, the decision we are facing maybe at different levels such as, we may want to decide which days do we want to choose for running our promotions. Or maybe at a lower level, we may want to decide which hours are the best for running our promotions. If you pause for a minute and think, you would realize that decisions and actions taken at the lower level, such as, changing discounts every hour rather than at a daily level, allows us more control on our business affairs. So you may be tempted to go down even further. Why not change discounts every minute? But using this dramatic example, I hope I am making this point clear that when you start going down further and further, taking business actions becomes operationally challenging. Managing discounts and changing it every minute is an operational nightmare and possibly a customer crisis too. Thus, as business analyst, we must choose the right level of analysis to maintain a fine balance between the control and operational feasibility. Through these last two videos, I hope I have been able to accomplish one goal, that, you can appreciate how closely related are managerial decisions, level of analysis, and data structure. This appreciation will prove very useful during data preparation stage. Before we get to how to implement the data preparation steps, I want to mention one last thing. In your dataset, each unique row is called a unit of observation. For example, if each row in the dataset belongs to a unique customer, then, the unit of observation is customer. Given this level of data, we can perform analysis either at the customer level or at a higher level, such as customer segments, but not below it. You will practice more on this during your assignments, and those will help reinforce these ideas better.
