You Ask, I Answer: Predictive Analytics for Content Marketing?

In today's episode, Monina asks, how can predictive analytics make your content more effective? So predictive analytics is a branch of analytics that allows you to understand what is likely to happen. And it's third on the hierarchy of analytics. There's descriptive analytics, which is what happened, there's uh diagnostic analytics, why it happened, and the third on the hierarchy is predictive analytics, what's likely to happen based on the data we already have. Predictive analytics comes in two general flavors.

One is uh understanding and building a predictive model of what makes something work, uh what variables, what what uh data points makes something work. And the second branch is time series forecasting, which is predicting when something is likely to happen. Both of these techniques are really important for content marketing because they make it easier for you to focus on what's working and help you diagnose diagnose and understand the timing of when you want to do something. So let's tackle uh uh the the first one, predictive models. Imagine you have a series of blog posts, you have the URLs for those blog posts, and you have the number of shares on various social networks, and you have the uh number of clicks and you have uh searches and you have all these numerical data points, and at the end you have something like you know, website traffic to to that blog post.

Maybe that's your outcome. What of those other variables has the strongest mathematical relationship to the thing you care about, the the website uh traffic. You would run a series of mathematical techniques to uh essentially a very large regression model to understand the mathematical relationships between all these other things and the outcome you care about website traffic. So maybe it is Twitter shares, but it's also number of words. Maybe it's uh the grade level readability score, maybe it's the number of linking domains or the page authority, whatever the the numbers or combinations of numbers are, uh there may be a relationship to the outcome you care about.

Using uh machine learning and and statistical techniques, you can build a model that says out of these you know 40 variables we have access to, these four or five seem to have the most importance, seem to have the most likelihood to predict a high traffic blog post compared to all these others. And from then, using data science, the scientific method using data, you construct experiments. Okay, if we can if it's something under your control, like uh number of words, okay. What happens if I write a longer article? You know, do uh do your create your hypothesis, run your test, and understand, yep, writing a longer article gets me more traffic.

Or maybe it is the number of inbound links. Great, let's go pitch this article to you know some bloggers and and see if uh we can get more inbound links and see if that is the thing that drives our content forward. So that's uh predictor estimator importance. To do that, you need really good software. Um probably the easiest one to recommend for people who don't like coding or don't know how to code would be IBM Watson Studio.

Uh that one has a uh what's called an SPSS modeler of sort of a drag and drop visual interface to do this cut type of modeling. It is still you still need to know the math behind it, but at least you don't have to do the coding part. Um but that will take that series of of uh factors or variables and that known outcome and build that regression model to tell you try testing these things next. Now it's important to say it's not gonna tell you the reason why something works, it'll only tell you the the mathematical relationship. That's why the data science part is so important because without the scientific testing, you might make an assumption that yeah, number of words is a thing, when it may not be the thing, it may not be there may not be a causal relationship.

You have to diagnose that on your own, uh using the scientific method. The second branch of predictive analytics is time series forecasting. And this is when you use existing data and forecast it forward to help tune your content. So let's actually bring up an example here. This is uh our cheese of the week forecast, something we do at Trust Insights for fun, um, to demonstrate the technology.

And what you're looking at here are all these numbers of cheeses, these names of cheeses, and then the predicted search volume uh for those uh terms, uh, using a combination of uh SEO data and Google Trend data and things like that, blending it together to build this model. And what we're predicting is for this week, the w we come recording provolone cheese will be the top search cheese followed by jack cheese and American cheese. And then if you look out at the end of the chart on the far right, the week of March 8th, 2020, uh Oaxaca cheese will be the the the top cheese that week, followed by American cheese. If you were a cheese shop, right and you were you're a cheese blogger, um, you would create content appropriate to each of these cheeses uh during the week that they're likely to be heavily searched. You would, you know, you'd be promoting provolone cheeses, you know, five different ways to smoke provolone cheese or whatever.

Um so that you are aligned with what people are searching for, because one of the things that is important in and search algorithms is the freshness of content and the relevance. Well, if you have the right content, relevance, at the right time, freshness, you'll m be able to amplify the results as opposed to just if you're posting uh about say Oaxaca cheese now, you'd do okay. Right? It's not like you would do badly, but if you that content were ready, say like the week before March 8th, you would get that freshness pop as well as the relevance pop in search algorithms. So this is an example of using predictive analytics to time our content to make it relevant at the right times when our audiences are interested most.

And obviously you can use this for more than just you know SEO. You could send out emails that week, you could schedule social posts that week, you could run ads that week for again. If you were the cheese shop, you'd be uh doing it based on the the type of cheese searched. Now extend this to your business, extend this to to what your company does. And of course you can see the the immediate relevance of let's get our timing right for all of our content marketing and what we do and how we distribute our content.

So predictive analytics can make your content much, much more effective and help you get organized. You can see this this chart could if you have enough back data that's good quality, you can forecast forward about you know half as much. So if you have five years of back data, you can forecast forward, you know, uh two-ish years. Uh I technically am more conservative, although just try to not forecast twenty five percent forward of whatever back data I have. Um, but you can forecast really, really far forward and then build a content plan for you know weeks, months in advance.

And that helps you get organized, it helps you take away that stress of what am I going to blog about today, or you know, what are we going to put in our social channels today? You don't need to worry about that anymore. Use the data that is publicly available uh with data science and machine learning techniques to forecast and use as predictive analytics. So, uh two really good examples of how to make your content more effective. Uh, if you have follow up questions, leave them in the comments box below.

Subscribe to the YouTube channel and the newsletter. I'll talk to you soon. Take care. Want help solving your company's data analytics and digital marketing problems? Visit TrustInsights.ai today and let us know how we can help you.