Okay everyone, we are now going to look at customer satisfaction analysis, and of course start with introduction. Let's first think about, what are the various types of data that we can analyze when we think about customer satisfaction? Now, we've already talked about survey data, but that's a major way that companies already look at data to analyze customer satisfaction, so we will definitely look at survey data as a source of measurement of customer satisfaction. But here's another dataset that's been increasingly popular, which is social media data, that consumers are going every day to social media platforms in announcing their liking or disliking of various brands, products and services, and so that's another dataset that we can look at to measure customer satisfaction. Now, if your company also has online behavioral data, so for example, how people are interacting with your website or how people are interacting with your marketplace, how people are commenting on things that you might be selling online, this is all online behavioral data, and that's also another dataset that you can analyze to assess customer satisfaction. Now, this third point, online behavioral data, we won't talk too much in this class because that's probably proprietary data to your company, and so we won't have public datasets to be looking at. But I just want to identify that that is a major way to assess customer satisfaction with regards to datasets that you own. Then of course, both now and into the future, there will be many more different types of data to analyze with regards to customer satisfaction, but we'll focus on survey and social media data. Now, let's talk about survey data. Survey data, you can collect either via physical or online means. Back in the day it was mostly physical, but of course now it's mostly online. Now, we need to build survey instruments with psychological and analysis concerns in mind. Remember from the first module, there is psychological construct considerations as well as measurement issue considerations with regards to constructs, and we need to keep all that in mind. But this class is not a class in how to build the survey because that could be its very much own class. In fact, many people's whole careers are based on trying to figure out how to create a proper survey. I'll just say that survey instruments need to be really carefully created with psychological and analysis concerns in mind. But then after you have created your survey and you've collected your data, usually, you analyze surveys with statistical or machine learning methods, especially for numerical data. We'll talk a little bit about statistical methods in this class. Now, with regards to customer satisfaction surveys, it's typically measured by how satisfied someone is with regards to a brand, a product, or a service. You may even rate specific components of your brand, product, or service, to identify which aspects drive customer satisfaction more heavily than others. We will look at an airline dataset in this course, and you might rate how comfortable the seat is with regards to your airline experience. Or you might ask for how people rate how good the food was. There's various components with regards to your product or your service, and these can all be measured with regards to customer satisfaction surveys. Now, a really popular survey, and I wanted to identify this one, because it's really powerful, a type of survey called a Net Promoter Score, or you might have heard of it as an NPS survey. An NPS survey is attempting to get closer to attitude strength because it deals with behavior. Now, we've probably seen this everywhere, we just didn't know it was called a Net Promoter Score Survey, where you're just asked one question, how likely are you to recommend this service or this product or this company to a friend? It's probably something from 0-10 scale, whereas that positive 10 is very likely to recommend, and that zero is very unlikely to recommend. The numbers may be different, but you get the gist. It's this range from bad to good, very unlikely to very likely. We've seen this before. Now, why is this a very powerful survey, this NPS survey? Well, because some of the pros is that it's very simple to set up and very simple to use. Some places I've even seen where they have these as physical setups where you're leaving a bathroom in an airport, and in that experience, there's some mechanism that you push where you either say positive or negative and you could even give potential variation of how positive and how negative, and it's some form of this Net Promoter Score. There's some evidence that NPS surveys are really a good indicator of company growth and company satisfaction. It's also adaptable to different aspects of your company. You can do a Net Promoter Score for your brand, for your product, for your service, and so it's really simple, it's really nice. But some of the negatives is that it lacks contexts behind the scores. Because you're only collecting one score, you have no idea if you're catching the same thing for each person. Let's say you say to somebody how likely they're to recommend Apple Music to a friend. You don't know if they're recommending it based on the catalog of songs or how quality the sound quality is with regards to each one of the songs, it just lacks context. Also, usually, Net Promoter Scores are collected over time, you see these emails that are sent to you, and so usually it takes quite some time to get full results, as well as there's little ability to apply statistical checks. You don't know if over time people are using that instrument in different ways. Remember this issue of reliability. With other types of surveys where you're asking a ton of questions, you can statistically check question to question to see, are they gaming the survey? Are they taking the survey seriously? Whereas regards to NPS, you have no idea because it's just one score. But what's really nice about the Net Promoter Score is it's getting closer to this concept of attitudes strength. Now, we also have this dataset that is social media. I mentioned before that people are more and more going on to social media platforms and talking about companies, brands, products, services. We can use this data to measure customer satisfaction. Now, usually you can click social media data via social media platform APIs, application programming interfaces, and we won't talk really at all about APIs in this class because that's not the focus of this class. But if you wanted to go to a social media platform and get data, you're likely going to have to interact with what is called an API. For satisfaction analysis, once you have data, which we'll give you some data in this class and tools to access that data without APIs, you can analyze that data using psychological texts models, and we'll talk about that in this class. For influence analysis, you can analyze using network analysis techniques, which we will also talk about in this class. For content analysis, you can analyze using text-mining, image processing, and or video processing techniques. Now, we won't talk about image processing or video processing techniques in this class, but we will definitely talk about text mining techniques to get at customer satisfaction.
