Speaker 1: Choice-based conjoint, or CBC, is a research tool used to optimize product features and price, learning what's important to buyers and for predicting what they'll choose. CBC mimics the real-life scenario where a buyer often compares multiple available products, considers what each option offers, and then makes a final choice. For example, I may need to buy a new TV. Some basic features of a TV are brand, screen size, screen type, and price. I would compare those details across different TVs offered in the market, trade off the features vs. price, and make a choice. Am I okay with an LED, or should I spend an extra $1,000 to have OLED? Maybe I could get OLED if I went with a cheaper brand. We all face situations like this, whether it's buying a new type of cereal, comparing health insurance plans, or planning a vacation. The best way to ask respondents to give us data for optimizing features and price is to mimic this experience. Choice-based conjoint questionnaires mimic real-world choice decisions just like this. Respondents are presented with several choice tasks made up of a few product concepts. They compare the available concepts and decide which they would choose. In this video, we'll go over how to easily add and configure a CBC exercise in Discover, show how the exercise appears to respondents, and then view the results. Now that we have a better understanding of what a choice-based conjoint is, let's make a CBC exercise. To start, we'll add the exercise to our survey by clicking the Add button, then selecting Choice-Based Conjoint in the menu. The exercise is then inserted into the survey. We'll use the TV scenario as our guide, assuming we are on the other side, as the store owner deciding which TVs we should sell in our store. First, we'll enter into our question, text, Out of all the TV options below, which would you most likely purchase? Next, you'll see there is a series of attributes where each attribute has two or more levels. Attributes are a feature category like brand, screen size, and price. Each attribute then has levels which are variations for that specific category. Our attribute brand can include Samsung, LG, and Sony. At a minimum, the software requires two attributes and two levels for each. We don't want to exhaust our respondents, so it's best practice to have no more than about 6-8 attributes and 20-30 total levels. So much also depends on the length of the level text and how much attention you think your respondents can pay to the exercise. We'll add our attributes by entering them one by one or pasting them from somewhere else like a Word document. After our attributes are created, we'll then enter the levels for each attribute. We can again type them out or paste in a list. I'll go ahead and paste in mine. Now that we've added in our attributes and levels, we're almost done. It's that easy. Over to the right, we'll see a preview of our exercise and see that it includes the attributes and levels we've specified. Before we publish our survey though, we'll verify a few more settings. We'll open the settings menu by clicking the gear icon. On the general tab, there's this setting for a none option. In most cases, it's best practice to have this setting turned on. If the respondent doesn't like any of the TV concepts, they can select I would not buy any of these TVs. The format tab allows us to customize the look of the exercise. We'll leave the task counter on so respondents know what set of concepts they are on. The total number of tasks is defined in the exercise design settings. We'll come back to that in a second. Prohibitions allow you to specify levels that should not appear together. For example, a 75-inch TV would never sell for $500. Using these prohibitions will prevent respondents from seeing an impossible pairing. We advise using prohibitions sparingly. Using many of these can have a negative impact on a respondent's preference score in analysis. Learn more about prohibitions in help documentation or in other videos. The last tab is the advanced tab. Here we'll find the exercise design settings. The software algorithm automatically generates the combinations of product features to show each respondent that will lead to a robust model for estimating the preference scores, also known as utilities. For our current lists of attributes and levels, the recommended design is specific to how many concepts we plan to show per CBC question as well as the number of items we included in our lists earlier on. If we were to change the number of concepts we'll show, or if we change the number of attributes or levels, the recommended number of tasks would be adjusted accordingly. If you want to override the recommended number of tasks or concepts per task, simply flip this toggle switch and enter in custom values. Remember that if you reduce the number of recommended tasks or concepts per task, your results may have less precision. If you are interested in learning more about what makes a choice-based conjoint exercise design, then check out our detailed documentation and additional videos. Now that we've verified our settings, we'll publish our survey and test it out. When taking the survey, respondents are shown several TV configurations or product concepts with variations of brand, screen size, screen type, and price. They then can select which TV they would most likely purchase, or none of them, if there isn't an option they like. They'll go through several tasks making these comparisons. In the end, we'll receive data that will help us understand what features are most important to our respondents, and we'll be able to predict how they would choose given different marketplace scenarios of available TVs at different prices. Now that we've collected a few data records, let's look at the results. On the Analysis tab, we can navigate to our exercise on the left. After we've collected new records, the software runs hierarchical Bayes estimation to calculate the utility scores. It can take a minute or more. The charts that appear show all our attributes and levels, and how each level on average is preferred to another. Combining the highest performing levels of each attribute, we could assume the product that may perform better on average than any other would be an $800, 75-inch LG TV with an OLED screen. Of course, there is more that needs to be considered in analyzing the results, but this helps to give us a quick insight. Discover also has a market simulator that helps us to conduct valuable what-if simulation predictions to better understand the effect that attributes and levels have on market choices. In the simulator, you can create several products and use the results from the study to see how well each one may perform in terms of buyer choice relative to another. In this quick example, I've created a few products. This chart shows how they may compete against each other in the real world. The simulator is the place where our CBC data becomes most useful. To make the most out of the simulator, learn about best practices and how to interpret the data and other materials on our website. Choice-based conjoint is an extremely powerful tool in helping us learn what's important to buyers and what drives their choice decisions. Learn more about how you can apply this method to your research through further resources available on our website.