6.1 What process do we follow to make progress? - Video Tutorials & Practice Problems
Video duration:
14m
Play a video:
<v ->As you're working with marketing analytics,</v> there are certain processes that can help you make progress in solving your analytics problems and in being able to improve the performance of your marketing based on analytics. And so, if we think a little bit about how the product management life cycle works, a lot of projects are done this way, so products are usually done like this. But we also see other projects that are done this way, marketing projects, where you might plan something upfront, design it, develop it, launch it, and you haven't really experimented with it. This process is called the waterfall process because of the way it looks in this picture. And the traditional way that waterfall processes work is for different people to be in charge of it along the way. So each person is kind of the expert for their step, and we do a lot of documentation. And now you might think this doesn't apply to marketing and applies only to products, but it actually applies to marketing too. If for example, you have a project where you're going to create 20 different videos for YouTube, the waterfall way of doing it would be to decide what all 20 videos would be about, start planning out what order you're going to do with them, design them, maybe by writing the script, then do the development by shooting all 20 of the videos, then you launch them all at once and you see how they do. That would be the waterfall way of doing it. But, there's actually a different way of doing it that might help you to make a lot more progress, and that didn't require as much planning and as much documentation, and certainly didn't require handoffs between all sorts of different steps. Because you could imagine that the person in charge of the project plan might be the project manager, but the person who actually wrote the scripts might be some kind of editor or author. And so, what else could you do? Well, there's an idea called agile development. And what you might wanna do is to think about using your agile development process to do your marketing projects. And the reason is because as you're going, you actually learn more about what's working and what's not, and it helps you adjust so that your marketing becomes more effective. So rather than follow that waterfall process, if you were trying to do those videos, maybe what you would do is you would say, hey, we know we're gonna do 20 videos, maybe we'll come up with some topics for a few of them, but then we're only gonna make a couple of them. Let's actually make two videos in a couple of weeks, and then let's try to launch them and see how they work. Because if they don't work that well, maybe we wanna change what we do for the next two, so that by the time we finished the 20 videos, maybe 10 or 15 of them are really good, and only a few of them were not that good, whereas if we did all 20 at once, well, now we've got a problem that we've got 20 videos that all were original idea, maybe none of them work that well. And so agile development comes from software development, so the idea is that you're constantly delivering things into production and testing with users to see if it works. There's all sorts of different names for it, you don't really care about much of that as a marketer, but you do care about the thought process, the thought processes for you to get things out there in front of your customer as quickly as possible, because that's what's going to allow you to test to see if it's working or not. And if it's not working, you adjust your course. And so, you can imagine that using analytics is a critical part of this process. And so using an agile process for your marketing is much more likely to help you make progress with your marketing because you're able to use the analytics to adjust as you're executing the project, rather than doing everything upfront and having your analytics be something after the fact that just tells you whether it's working or not. If you're doing it while the project is underway, you're actually able to adjust your course so that your marketing will work better. So I can actually prove to you that agile is easier than waterfall. And the way I can prove it to you is that waterfall is more like baking and agile is like making soup. So what do we mean by that? Well, with waterfall, you have to know exactly what you're baking upfront, you have to know exactly what you're doing with everything, measure everything perfectly, and nothing's done until the end. Only when you had all 20 videos did you have your project done and then you launched. Well, that's not how you did it with agile, agile is more like making soup, where what you're doing is you're actually testing it all along the way. You can taste it and you can see whether the soup tastes good, and you'd say, oh boy, we need to add a lot more salt. And so that helps you to make the soup better. And so it's much easier to do things in that agile methodology, but it's also something that will make the quality of the results work out a lot better because you're tasting along the way. And you can think of analytics as your customers telling you how it tastes. And so by launching a couple of those videos and seeing how customers responded to it, you're actually letting your customers taste the soup before you're finished with the project. And this is why it's easier to do agile than it is to do waterfall. And I can actually prove that to you with a simple example, which is it's easier to make soup than to bake. I bet all of you know people that are good cooks, but they don't like to bake because baking is just too hard, because we could make soup the same way we bake. We could put everything into a pot, simmer it for four hours and just ladle it into bowls on the table without ever tasting it, but it's harder to do that way, and the soup wouldn't taste as good. So anytime you have a chance to do your marketing in an agile way, that will allow you to have a better quality of product, when you're done, your marketing will work better, and it allows you to use your analytics to improve all along the way. So how do we do this data-driven marketing process that uses this agile approach for our marketing? Well, we start by defining objectives for our measurements, then we do some experiments, so that's what those first couple of videos were. Then we test them, we launched those videos and use analytics to tell us whether they're working or not. And we monitor the performance, then we come back around and do it over and over again until we have a set of videos that we really like, and that can be stuff that we keep in production all the way. And so this data-driven process that uses analytics to constantly inform our decisions, this is what improves your marketing more than anything else. So let's think about the specific steps in your data process that helps you put that analytics together. So how does your data process work, not just your development process for your marketing? So the first is to gather the data. So how do you collect the data from sources inside your organization, but maybe from outside too? Maybe you're pulling in data from Facebook insights or Google Search Console, maybe you're bringing in data from all sorts of places that you're going to aggregate together, gathered together and be able to use to solve your marketing problems or to assess your marketing performance. Second step is to synthesize. So you need to take that data and put it together in such a way that you can relate the data to each other. So how do you get it into some type of common structure that you can export to a data lake or to some other type of a database where you have a common schema? Third step is to analyze, and we highlight this step because this is by far the most important one. Yes, you have to do these other steps, but it's the analysis that's actually gonna help you solve the problem. If your analysis isn't done properly, it doesn't matter how you've done any of these other steps. Each step is important, but the analysis is a critical one and would probably be different for every marketing problem that you go after. And you can not only use algorithms, you can use pivot tables, you can use AI models, but what you wanna do is to really identify what the hidden failures are that are breaking your marketing. What are the things that you can do to make things better? After you've done the analysis, there's two last steps, one is to visualize the data, that's what helps it make sense to people, what gives you insights, and what's helped people to take action? The last one is to actually automate things. So can you take the data and use automatic actions by the computer to do something based on the data? So can you set up content recommendation the way Amazon does, where you're actually recommending products based on which products they've looked at? That's an example of using your data and your analysis and automating it so the computer is actually taking action. So once the analysis is done, the visualizing and the automation, or what humans do to take action and what computers do to take action. But that analyze step is all important. So you can use some kind of vendor dashboard, or you can pour it into your own, you can have machines taking action. Those are relatively simple steps compared to the analysis step. The critical point is for you to use AI models or predictive analytics or other techniques to actually solve the problem. Once you know the problems being solved through your analysis, then allowing humans to take action from those visualizations, or to have computers take action will be relatively simple. So I mentioned artificial intelligence, so machine learning is a key technique in artificial intelligence, and it can often be a game changer in how you extract insights. Because what it's doing is it's looking at data and predicting outcomes. And so the process that you would use for that is to identify the training data, train your AI model, deploy the model, and then start to assess it. Start to assess whether it's working well or not. And you can do this in a test mode, and do this over and over again and keep adjusting your model until you get to something that you can deploy in production. And there are a lot of different marketing problems that this works for. And so if you think about the different use cases that a machine learning algorithm could help with, it helps you tag your content by topic or by industry, it can help you predict different outcomes like bounce rates or exit rates, it can help you predict whether user tasks will be completed or what kind of content you should be recommending in real time. Any of those things are things that can be automated using a machine learning approach. So let's use an example of that. So suppose you want it to identify which keywords you're missing from your search engine optimization and paid search programs. Well, one of the things you could do is you could collect all of the keyword for those two programs, but also from your site search capability. So where are all the keywords that people are entering? So some of them you're already targeting in your organic search and your paid search, but some of them might be things that people are entering on your website and your site search. So, how do you collect all that together? Then, what do you do to synthesize it? Where are you gonna match keywords together across those sources? Now, some of them are simple because they exactly match each other, but you also wanna think about the fact that when people are entering keywords in your site search engine, well, those keywords, they're not gonna be entering your brand name or your company name the way they might when they're entering them into Google. And so you might wanna strip the brand names out and have more matching that's being done without the brand names for those unbranded keywords. Then what kind of analysis do you wanna do in that third step? Well, you wanna aggregate the data together, pivot the keyword data across the resources, and that will show you what types of things you might be targeting in one of the campaigns, but not in another. It might also show you site search keywords that have high volume associated with it that you're not targeting in either of your campaigns. You can visualize that on a dashboard. And so you can show that so that you can have humans make decisions of what to add, or you maybe you could automate things. Maybe what you could do is just say, hey, any keywords that's being used over a certain amount of times in a month, let's just automatically put that into play as something that we wanna create an ad for. Let's send it to the agency and have them put an ad together for our paid search. Any of these things are things that you could do. So this is an example of how you can solve an analytics problem with a particular process that walks you through these five steps.