In this video, we are going to look at how to simulate a coffee shop. In particular, we are going to talk about how to model the arrival of customers or how many customers arrive and what revenues are generated in a day and over a longer period of time. We're going to look at modeling this problem at a variety of different levels. These different levels depend upon the data availability as well as the requirements. By requirements, I mean the questions that we need to answer for the business owner or the coffee shop owner. Do this through extensive labs that build progressively in its sophistication of simulation modeling. During these labs, we're going to talk about the state of the art practices that are involved in Monte-Carlo simulation modelling. Let us assume that we are hired by a coffee shop owner to build a simulation model to understand or maybe even predict the revenue for a coffee shop based on information regarding its past performance when we have to develop the simulation model. The first thing as a modeler, you should ask what data is available because the level of model and the type of model that you develop will depend upon that. So the data usually can be available, in various granularities. For example, we may have average number of customers per day, week, or month, or just simply minimum and maximum number of customers over a period of time. We may have average sales available per day, week or month again, or just the minimum or maximum values of those. Or we may be lucky and we might have even better information such as hourly arrival of customers and sales data. And maybe even the data on how much time consumers wait during the busy time. Now the model we develop should be dependent upon the question that we are trying to answer. There is no point in developing a very sophisticated model If the question that we have to answer is at a very aggregate level. On the other hand, some type of models simply cannot answer the questions if they are at process or details level. So questions such as daily range of revenue, range of Revenue given number of customers daily revenue versus monthly revenue. Profits, versus just the revenue, waiting time of customers or effective waiting times or customer spend, or whether a customer stays in the shop or not. These questions may require different type of modelling and we may have to collect different types of data to answer these questions. So again, modeling details for each question require either different levels of data or necessity to make assumptions. And sometimes both. For some questions, we may need to acquire additional information than may be available. This might be collected through surveys or questionnaires or going and spending some time in the business and collecting the information firsthand. Once we do the modeling, our job is not over. we need to do what is called sensitivity analysis. The need for sensitivity analysis in Monte-Carlo simulations or any type of simulation arises due to several reasons. First of all, it may be that any assumptions that we have made in the model, these assumptions may be about the actual numbers or the shape of the distributions that we're using to model uncertainty. Furthermore, the inherent variability that we are modelling during the simulation model to capture the uncertainty produces non-unique results. That means the results may be different every time a model is run again. Now, the sensitivity analysis in Monte-Carlo simulation essentially involves doing a lot of replications. And to understand the stability and volatility in the outcomes of interest. By collecting this information, we can develop more confidence in the results or outcomes. And we can tell which outcomes might be a little susceptible or very, very sensitive to changes in the patterns, changes in the input patterns. So in the next set of videos, we're going to simulate coffee shop revenues. We'll start with single uncertainty, which is just simply about number of customers that are arriving, we'll do the sensitivity analysis of this data that we generate. And then we'll answer some business questions. And then we'll see which business questions can we not answer. We'll then go on to simulate with multiple uncertainty, where uncertainty is not just in the number of customers, but it may be also how much each customer, for example, spends. So we'll see how we can model that. What are the benefits of modeling that? And we'll do a comparison of results and sensitivity analysis with that data.
