In case you haven’t heard, B Squared is hosting the first annual HELLO Conference on March 28, 2019.
We’re super excited about this!
We’re excited not just because we’re hosting our first ever B Squared event, but because we have an incredible line-up with seasoned pros like Katie Robbert of Trust Insights.
Katie was kind enough to sit down for a session of B2TV with us ahead of the HELLO Conference where she discussed the different types of predictive analytics and what this means for marketers.
If you’re already feeling confused and overwhelmed by the concept of predictive analytics, have no fear!
Katie did an awesome job of breaking down it down so we all can understand not just how it works, but how it can be of benefit to us.
If you’re interested in learning more about AI and predictive analytics, check out our interview with Katie.
Defining Predictive Analytics
B Squared: How do we break down predictive analytics in a way that most marketers could understand? The Layman marketer. How do we define predictive analytics?
Katie Robbert: There’s two different types of predictive analytics that we use in marketing.
- One is Driver Analysis.
Driver analysis at a high level is taking a lot of data, basically variables, so if you’re looking at social media, for example, you have a lot of variables around likes, retweets, comments, length of post, day of post, week of post, and so on. There’s a lot of different variables that you can look at. With all of that data, what you want to do as a marketer is figure out what combination of things is driving a higher engagement. Driver analysis can actually predict down, of all the variables, here are the top three that are a priority to your specific audience. Here’s what they care about.
So it might be the length of the post. When the post contains an image and not a video, and then when it’s posted at, for example, 2:52 on Tuesday afternoon. Those might not be things that you’re thinking of, but the driver analysis can winnow down the best predicted combination of things that will get you the goal you’re looking for.
- The other type of predictive analysis that we use in marketing is Time Series.
This is the one that people are probably more familiar with.
Time series analysis takes any data set that’s time-stamped consistently with a date.
We often use Google Trends data because it’s very clean and very consistent. What it does is it uses statistics to say, of all of this historical data, if I project it forward, here is likely what’s to happen with these trends based on what has happened historically. That’s a lot of what I’ll be talking about it at The HELLO Conference in a few weeks!
Using Predictive Analytics to Build a Better Strategy
B Squared: So, I want to go back to Driver Analytics for a second. Because what you said is so interesting. We’re constantly being, as marketers, barraged by “Best Practices” all the time like “your content should only be 80 characters on Facebook”, or “You HAVE to post on Wednesdays at 3 pm because that’s when posts perform the best!”
Personally, I hate best practices because it’s putting out a blanket statement onto your own industry or product or service. So, does the driver analysis that you have done show that everybody has their own unique best practices?
Katie: That’s absolutely correct, because not every marketer, not every company has the exact same audience. If everything was created equal, then yes, you could say this is the best practice for doing it. Really, those best practices should be guidelines and then you need to look at your specific goals.
Your goals might be different from what the guy down the street has for a goal, you might be posting the exact same thing but what you care about is different from what he cares about so you can’t say that you have to do everything the exact same way because getting the behavior that you want out of your audience isn’t going to take the same type of techniques and that’s where driver analysis can really start to determine based on your goals for B Squared, or my goals for Trust Insights what we need to do to get the best out of our audience
B Squared: So, give me a simple example of how you would perform that Driver Analysis with some simple data sets. What would that look like?
Katie: So, what we would do is we would extract all of the different variables that we could from a Facebook page and that might be what type of thing is in the cover photo, how many reviews they have, how much interaction they typically get on a post. You really start to pick apart all of the different variables that you can pull out of any one page, and then what you would do is, it sort of becomes like a very large spreadsheet and unwieldy thing with hundreds of variables, and then through the magic of statistics and math, you run the proprietary algorithm that we have created, and it would then start to tell you based on what your goal is, what you care about, what specific things you need to do more of.
So maybe you need to do more posting on Tuesdays, or you need to do more longtail content, or you need to have shorter titles that include a hashtag, or something along those lines. So those would be the types of things that you would be able to determine through driver analysis.
Another really good concrete example is, if you’re looking at your CRM data, and you want to know what things influence recency and frequency and higher value then you can start to cluster your customers into buckets of frequent purchasers or repeat customers, and then maybe you have customers who make a one time purchase or make multiple purchases but they’re smaller dollar amounts and then you start to pick apart the variables that comprise those customer profiles. Then you can start to figure out the profile of the type of customer that you need more of to get the higher value, more frequent customer.
B Squared: Could that kind of analysis help you determine “x type of content brings in more followers, x type of content brings in more engagement, x type of content drives more link clicks. To bring in the answers for social KPIs?
Katie: It does. So you would run a couple of different analyses, based on each individual goal so that you could figure out “okay my goal is more link clicks” or “my goal is longer read time” or whatever the goal is, and then it will start to determine what combination of things drives the behavior that you’re after.
B Squared: Awesome. When we do our content for our clients we do a human report called the “SeeSaw” report. I’m going to give you a scenario, Katie, and you tell me if I’m if I’m on the right track here.
So we do the SeeSaw report every week where we look at the top performing three posts on every platform the client has and the bottom three performing posts. So its trends and patterns, essentially, what isn’t going well inside of the content and what is going well. Deciphered from a human and then we come up with “Hey, moving forward we should try this type of content or do less of this type of content”.
My guess is, you could use some of the algorithms or software that you guys have to let the machines kind of take that over and take the human work out of it for trend and pattern discovery when it comes to content.
Katie: Yep, that’s exactly the problem we’re trying to help automate, a lot of those redundant tasks. So, if you do the same thing week over week, even if it’s pulling different links, that is something we feel is really right for that automation and allowing the machine learning to do the analysis, so that you can then focus on the deeper insights and coming up with better actions, and that’s really, at the crux of it, the goal of kind of company we are trying to build.
B Squared: I love it. That makes so much sense because the machine is really supporting the human at that point. The machine is doing kind of that heavy lifting, that really mundane type of tasking where Carrie can take what the machine spits out and, say “oh wow, based on this information, let me put together five different ideas for how the client can move forward with the good performing content and that’s where the human comes in and gets involved with the machine’s help.
We’re not taking the human out we’re just having the machine help the human be better.
Katie: That’s exactly it. And I believe that’s exactly what you are trying to promote at the HELLO Conference, and that’s what is the basis of my talk and Chris’s talk is. It’s about how artificial intelligence and machine learning really supplements the human because there are limitations to what you can do with machine learning and artificial intelligence and primarily it’s that deep, critical creative thinking that the machines can’t do at this time.
B Squared: Can I invent a super cheesy catchphrase? “Robots make the humans Super Human!”
Predictive Analytics and Efficiency
B Squared: Digging a little bit deeper, for those people who are not “there” yet. Why is AI so important to social media or digital marketing and adverting? Why is AI so important to what we do? How is it changing what we do?
Katie: The amount of data that’s available at any given moment is so overwhelming.
There’s too much information available for any one human to just be able to process and make sense of. So you need that machine assistance in order to make sense of it, to clean it up, to make it something that you can do or take an action with, to get deeper insights from. It’s really the sheer volume of information.
The other thing is that the machine, if programmed correctly, takes the error out of the analysis. It will do the analysis the same way every time based on what you want it to do because we, as humans, get fatigued.
As we get tired, we might transpose a number, machines don’t do that and so you’re not going to worry about your machine calling in sick when a big critical client report is due, which has happened to all of us.
I you need the machine to work, 24/7, 365, it’s going to be able to do that and it’s going to be able to give you the analysis you’re after.
Why predictive analytics is so important from my standpoint is that the challenge with reporting and analysis is it’s looking historically behind you.
You can’t change what has happened. You can only say, all right, how do I learn and grow from it? Or how could I do something different? What we’ve seen in our experience is 99% of the time people aren’t even doing that. Companies aren’t even looking at what happened historically, they’re just making decisions and reacting and moving forward.
The goal of using something like a predictive analysis, specifically, time series, is to be able to plan out a content calendar and to be able to know what’s going to happen to be more prepared so that you’re not so reactive. For social media marketers, for example, you’re producing content, you’re looking for when hashtags might be trending, you’re looking for what topics people are going to care about. Using a time series predictive calendar is going to allow you to do all of those things.
For the full interview with Katie and to learn more about predictive analytics, check out our B2TV session!
Latest posts by Brooke B. Sellas (see all)
- 30+ Effective Ways To Enhance Social Media Customer Service - February 21, 2024
- Leveraging (and LOVING!) Analytics for Smarter Digital Engagement - February 14, 2024
- Building an Effective Social Media Customer Service Team - February 7, 2024