Welcome back. In this video, I will introduce you to a process to perform conjoint analysis. Remember, the purpose of conjoint analysis is to determine how useful various attributes are to consumers. To put this into a business scenario, we're going to look at how conjoint analysis might help you design a flat panel TV. After this lesson, you will be able to describe the conjoint analysis process, list the steps in correct order and explain what takes place in each step. I have broken this process down into five steps. Step 1 is to identify important attributes. Step 2 is to select the attribute levels. Step 3 is to construct product profiles. Step 4 is to collect consumers' preferences. And step 5 is to analyze the data. Let's look at each step in turn. Step one is to identify important attributes. You would start by thinking about the important attributes that go into deciding when a consumer wants to buy a flat panel TV. You might think about projection technologies, screen size, refresh rate, the brand name and the price. Your goal is to identify the attributes that are important to your target market. Step two is to select the attribute levels. Within each of these attributes that we just discussed, as a manufacturer of a flat panel TV, you might want to think about the different levels that could exist. These levels that we spoke about are defined in a very particular way. For example, with respect to projection technology, you might think about levels as being OLED or LCD or LED. With respect to screen size, it could be a 46 inch TV, a 52 inch TV or a 65 inch TV. In our case, we have also provided levels for the other attributes like refresh rate, brand name and price. The refresh rate can be thought of as 60 or 120 hertz. The brand names that we're considering would be Sony, Sharp or Samsung, and the price is either around $1,000 or $1,500. Step three is to construct product profiles. If you go back to the attributes and the levels that we have just discussed you would see that there are three tech attributes, three sizes, two speeds of refresh rates, three brand and two prices. All in all, that's about a 108 profiles, which is way too many. Given a large enough numbers of consumers, 108 profiles might be okay. However, at times it could be too much, because you have to show consumers several of these profiles, so that you can get enough number of observations for the different trade offs that consumers will be making. So that you can identify which profiles are better for your particular products and why. In a situation like this, the thing that is often recommended is to reduce the number of attributes, the levels or both. In our case, we want to try and cut this down to about 24 full profiles. We do so by cutting down some attributes and some levels. What we'll do now is cut things down to three sizes, two refresh rates, two brands and two prices. Let me go a bit deeper here to illustrate what it looks like to build a product profile. I will use two flat panel TV as example. This is still part of step three, but I'm demonstrating what that looks like in practice. Consumers will presented with a product profile to compare to another similar product with different attributes. For example, the first profile called Profile 13, is a 52 inch TV, 60 hertz, has a Sony brand name and cost $1500. Lets say this profile is shown to consumers alongside another profile called Profile 9. The second product is a Sharp TV. The Sharp TV is 46 inches, 60 hertz and $1000. What we did with these profiles, you would go on and do for 24 different profiles based on the attributes and the attribute levels that you have selected. Step four is to collect consumers' preferences. When each of these profiles is presented to consumers, they're asked to rate their preferences on a consistent scale. Here we see that as a scale of 0 to 100. By having consumers compare products side-by-side with various attributes and rate their preferences, you can compare these ratings. Essentially, we're trying to relate these attributes and attribute levels to the ratings that consumers provide from different profiles. There are several approaches to collecting consumers' preferences. For example, you could do rankings, you could do ratings, you could do paired comparisons between the different products or you could have a choice based conjoint analysis. Whatever approach you use your aim is to get consumers in your target market to tell you which attributes and features matter to them and to place a measurable value on those products with those particular attributes. Analyzing data is step five. Going into step five, you've collected the data by collecting consumers' preferences. Once you collect the data, of course, you want to analyze it. There are several approaches available to analyze the data. The methodology you choose will vary depending on the approach used to collect the data. In the example I used comparing flat screen TVs, we collected ratings information. That lends itself to linear regression as an analysis method. Each of the other approaches we mentioned, rankings, paired comparisons, choice based conjoint analysis has different methodologies best used with that approach. So you would choose the appropriate methodology to analyze your data. One way you might analyze such data is using a regression where the dependent variable Y is the ratings or preferences of the individuals, and X is the independent variable that is the attribute. Once you do that, you can then obtain a table of coefficients. Now, I won't be going into how regression's actually done, there are several courses in Coursera itself, that you could check out that addresses this issue. Our focus is to see how the results of that analysis and the entire conjoint analysis process relates to your market research report. What you see in this table is the results obtained from the regression. You see that there is a column that we call coefficients. These coefficients essentially tell you how much a level of an attribute is worth. Now, remember, these are not complete attributes. These are parts of attributes or levels of attributes. Rather than calling them coefficients, we'll call them part-worths from now on. The question we intend to answer in your market research report is, how are these part-worths useful in making decisions in the marketing context? Let's first address why part-worths actually matter. For example, you see that a 65 inch screen is very highly valued. It's at about 10.75 utils. Utils are a measure of the value for that particular part-worth. Next is price, where the part-worth is -4.75 utils. After price, Sony brand name, and after that refresh rate are the most valuable. By looking at the magnitudes of these coefficients, we can rank all of the different levels of attributes, so that we know which attributes are more important than the other to our customers. Now, we can use these part-worths to calculate multiple things. For example, we can calculate the willingness to pay for a particular attribute. We do that by first looking at the two prices which were $1500 and $1000. The price difference between these two price levels is about $500. Now let's look at the price part-worth, which if you'll recall was about -4.75 utils. What we do next is relate these utils to the price savings, which was about $500. We can then obtain the value of one util, which is obtained by dividing $500 by just the magnitude of the utils, which is 4.75. This results in a singe util being worth about $105. Next, we can use this information to try and determine the willingness to pay for different attribute levels for that product. What we have done first is to try and determine how much one util is worth to an individual. We find that it's about $105. If you recall, the way we did that was we looked at the savings between two different price levels, and we looked at the part-worth of price which was 4.75 utils. And we divided the two, so we can obtain the value of one util of the coefficient of the part-worth we looked at. Let's say that we want to try to get at the willingness to pay for 65 inch screen. The way you would do that is to take the part-worth of the 65 inch screen and multiply it into the value of one unit of that part worth, or 1 util. You see that the willingness to pay for a 65 inch screen is about $1135 more than the baseline part-worth, which was the 42 inch screen. The willingness to pay for Sony name is about $360 above and beyond the Sharp brand name. Willingness to pay for 120 hertz refresh rate is about $237 more than the 60 hertz refresh rate. What our example showed is that conjoint analysis enabled us to look at the product features of a television and understand which features are most important or are most significant for consumers. And even determine how much each of these features is worth to consumers.