 Hello there! It's me, Monica Wahee, here with some more data science learning advice. Only today's advice actually comes from one of my learners, Lance Zeng. I learned he was my learner because he so nicely posted an article about my LinkedIn learning course, the Data Science of Experimental Design. I'll link you to Lance's post in the video description so you can read it yourself. But what was impressive to me, the author of the course, is how well Lance learned the lesson, which is that you actually don't need an A-B test when you think you do. You actually just need to optimize your system first. For example, here is his big quote, the goal of an experiment is to compare the effects of different conditions for an event of interest and provide insight into inherent cause-and-effect relationships. Okay, so what that implies is that you are in an environment where all the potential causes are controlled and you are just changing one little thing and seeing if that one little change makes a big enough difference in the outcome to warrant making it part of our routine business. In a way, it's a little like trying to make a machine learning algorithm that beats an existing algorithm. Here, Lance shows us a diagram. It's nice. It shows us just changing one thing, the color of the cup and the picture. If we change that from red to blue, do we sell twice as much? Sounds simple, right? But the issue is comparability. You need to be looking at comparable outcomes from each condition, like you can't be expecting a click in one condition and a sale in another. And you need comparable conditions. One condition can't be a way better deal than the other, or you'll obviously have a higher conversion rate. And also, you have to use comparable timeframes to compare rates. Oh, and Lance catches on to that too. How important rates are compared to counts. He points out the benefits of using rates and also gets into some details I cover in my course about choosing a denominator. One of my learners asked for a concrete use case example. So this year in 2022, I'm going to try to build my YouTube channel following. So if you like this video so far, please subscribe to my channel, actually click the like button, or make a comment. That's one thing I learned so far. Ask for subscribers. And how did I learn that? I took this online course on LinkedIn learning by Anson Alexander called Marketing and Monetizing on YouTube. It was a very good course and I recommend it. I'll link you to it in the video description. But I pulled out some concepts from that course that I wanted to highlight in terms of how they intersect with my course on experimental design that Lance was talking about in this article. So the main thing we are talking about in marketing is conversions and conversion rates. So Anson mentions these possible YouTube conversions. These are conversions I could use to track my YouTube channel growth. I could track number of subscribers, number of views of a certain video or of all my videos, and also the amount of money I'm making on monetizing. So let's say I do all that. The issue is I have limited time. So what strategies worked best for me? Like if I had to jump in views or jump in subscribers. What caused it? I want to know because I want to like do it again. But what did I do right? And that actually is an issue that Anson covers somewhat fleetingly in the course. First, let me say that it's in chapter one that he discusses the different conversions I mentioned. But he discusses them under goals. But these goals can also be personal goals like post so many videos per week. So not all of them are conversions per se. But then in chapter four, he gets to the issue of how do you figure out what works? He is smart in that he recommends varying videos to attract subscribers and seeing what works. For example, I have a lot of videos demonstrating how to use R and SAS. So basically programming demonstrations. Those videos are pretty short. But I also have videos that are essentially lectures on topics such as how to do healthcare quality projects. Those are long form. By doing these different types of videos, I can gauge who my audience really is and what they really like. If you take the YouTube marketing course, which is called Marketing and Monetizing on YouTube and is on LinkedIn Learning, you will see the author just mentions a B testing or experimental design in chapter four. That's because it's such a huge topic. He couldn't really get into it in his course. But I can and I do. So thank you, Lance, for that write up about my course, the Data Science of Experimental Design. And to all of you watching who want to get into experimental testing, I strongly encourage you to take my course. Since Lance was a happy customer, I think you will be too. And if you do take the experimental design course or the other YouTube marketing course, please let us know what you learned. Leave a comment on this video. And thanks for watching. I hope you have a nice rest of your day.