Demand decomposition is a well-known proven method for forecasting demand. We shall demonstrate this method in the context of two examples. The first example we shall consider is an example of estimating demand for car-sharing in Philadelphia. The next one is demand for toothbrushes in India. Let's start with the demand for Uber or car-sharing in Philadelphia. In demand decomposition exercises, we invariably start at the highest level, and here we'll start with the population of Philadelphia. The next step is to figure out those who need transportation. If we are specific, maybe it's transportation to work or it could be more general transportation. Next step would be to multiply that with those who prefer personal transportation over public transit, and then multiply that with those who do not own the car. Then those who will prefer Uber over other means such as taxi. If you multiply Philadelphia population times those who need transportation times those who prefer personal transportation over public transport times those who do not own their own car times those who prefer Uber or other means such as taxi, you get an estimate for demand for Uber or ride-sharing in Philadelphia. Again, each of these multiplications is by a fraction. If you look back in this application, we had two steps where we use the word or phrase, those who prefer A or B. How do we find this out? One way to do this is to ask people's preferences or intentions. However, we know from previous research that people often do not report their intentions accurately. Even if they report them accurately, oftentimes they don't behave according to the intentions that they had reported. This is a complex problem, but it's a solvable problem, and there are methods available to make intentions a better predictor of behavior. There is a lot of work and research done on that, but this is beyond the scope of this specific module. But have the confidence that you can, using these methods, make intentions a better predictor of behavior. Now, let's look at a second example which is a little bit more detailed. Let's think of a company wanting to better understand demand potential for toothbrushes in India. Because it probably wants to market a different a toothbrush. Again, demand decomposition model is a useful method in this particular example as well. Again, we start at the highest level. What is the population of India? This number is readily available. Next, we look at how many people brush their teeth with some form of device or another. Some people use neem twigs in India and others use a regular toothbrush. These data are usually available from studies published by the government and other organizations such as the ADA. The next step will be to look at the frequency of brushing. Frequency of brushing varies in every country. Not everyone brushes two times a day. Some people brush once a day. Some people brush less frequently, even though we are supposed to brush two times a day. But the variance is probably more in India. These data will come from surveys that you have to conduct or have already been conducted. Finally, after how many sessions of brushing is a brush replaced on average? There is a variance here as well across the population, and this can also be obtained by surveys or observation. The company in question did all this and came up with a very nice demand model. Starting with the Indian population at about 1.2 billion. Those who brush with some device or another, close to 50 percent, and of those who brush, how many of them use traditional means versus how many of them use a store-bought brush that we know as a toothbrush. Then number of brushing sessions per year, how many times do they brush a year? The data suggested that on average, people brushed once a day, so 29 brushings in a month times 12 months. Then the next question that they asked was, how often do people replace a brush? On average in India, based on the studies, it was after 160 of brushings. That's once every five months or so. Then that leads to when you multiply the total population with the fraction who brushed times fraction who use store-bought brushes, number of times they brush and how often they replaced the brush, you get an estimate of demand for toothbrushes in India. That number, you should check to validate your analysis whether it is close to the actual demand today. Now the next question is, what can you do to increase the demand? Well, that's what these models are very good at. Maybe you can increase demand. You can't change the population of India, or at least that's not under your control. It change on its own. Maybe you want to persuade people to press for the first time. But that will require a very strong health message on the importance of dental care. Alternatively, you may want to target people who use neem twigs and convert them to toothbrushes. Or not worry about that, just focus on the incidence of brushing. People on average are brushing once a day. Maybe you want to convince them that brushing twice a day is more important. Or alternatively, you could convince people to replace their brushes more often because keeping the brush for a long time is not as healthy. What does demand decomposition give us? It gives us an estimate of the demand. But more importantly, it tells us how to increase the demand. What are the levers we have under our control? Based on this model, we can come up with ideas on how to enhance the demand for toothbrushes in India. We could persuade more people to brush. We could urge people who use twigs to shift to a toothbrush. Maybe an entrepreneur could come up with an inexpensive toothbrush that will convince people who are using twigs to use toothbrushes, or one could urge users to replace their brushes more often. Demand decomposition allows us to also identify levers that can be used to increase demand for our product, and also identify how you might position your product in the market. In summary, forecasting demand is probably one of the most critical aspects of launching a new venture. We started this in three different ways. First, by using the ACCORD model developed by Professor Everett Rogers to examine whether or not our idea has a good chance of becoming successful. Then we discussed qualitative methods such as using expert opinions, benchmarking, and combining forecasts. Finally, we studied demand decomposition, often referred to as the chain rule. I urge you to apply one or more of these methods to your new idea. I wish you all success.