June 11, 2019
In this episode, we talk with author Jim Sterne about artificial intelligence in marketing. Jim wrote the book on the topic and shares his thoughts.
Christoph Trappe: Hello, everyone. It’s Christoph Trappe, chief content officer at Stamats. Thanks for joining us for another episode of our Stamats Insights podcasts. And today, I’m joined by Jim Sterne. He is the founder of the Marketing Analytics Summit and director of Digital Analytics Association. Jim, thank you for joining us today.
Jim Sterne: Thank you so much for inviting me. It’s a pleasure to be here.
Christoph: I’m so amazed how small of a world we now live in. So, the reason Jim and I connected is because I was actually – I had some Europe travel booked and I was flying back on Iberia. Now, if you don’t know, Iberia does have Wi-Fi, but they charge you by how much you use. So, you run through Wi-Fi really quickly. And I was looking for some books to read.
And one of the books that I actually bought is a hard copy of is Jim’s book, Artificial Intelligence for Marketing: Practical Applications. So, I spent my—I don’t want to say eight hours, I did take a nap. But I did learn quite a bit about artificial intelligence. So, thanks for writing that book and thanks for joining us today.
Jim: And thank you for buying a hard copy.
Christoph: People still do that. I just did. So, artificial intelligence, I mean the book was written a couple of years ago now, right?
Jim: Yeah. That’s correct.
Christoph: But it still seems very relevant to me and still something certainly I have to hop into more and more. But if you can get us started with—give us some overview of what do marketers have to know about artificial intelligence?
Jim: Well, when I write a book, and I started writing books back in 1995, about online marketing, I recognized that the topic was going to change quickly. So, I was careful to focus on what do you need to know that’s going to stay evergreen?
So, chapter 2 is, if you are a marketer, here’s what artificial intelligence and machine learning means so that you can have a conversation with a data scientist. Chapter 3 is, these are the problems that we’re trying to solve in marketing and the data that we have available so that a data scientist can now have a conversation with a marketing person. So, when you ask, what does a marketing person need to know right off the bat? The first thing is that artificial intelligence is this broad umbrella term that covers robotics and self-driving cars and natural language processing and computer vision, and this thing called “machine learning.” And I really wanted the book to be machine learning in marketing and the publisher said, ‘No, no, no. That’s not sexy enough.’
So, okay, fine. Artificial intelligence. AI, artificial intelligence at large is important for marketing people to understand because it’s ways we will communicate with customers. We will communicate through chat bots. We will communicate through social media algorithms. But machine learning is how we process the information we collect in order to decide who we should reach out to, what we should be saying, what the next best action might be. So, AI. Big term. Machine learning is a new kind of programming. That’s what I was trying to get across to marketing people. This is an introduction to artificial intelligence for marketers.
Christoph: And you know what I found really interesting is as you’re talking about machine learning, you made a comment in the book about if you have people who don’t even do a good job, you might as well have a machine who doesn’t do a good job or something like that. And what’s interesting about that is as the computers are learning how to react and how to do things, it’s very interesting because we’re all striving for perfection, but it’s almost a thing that doesn’t exist, I guess.
Jim: Well, one of the benefits of age is the recognition that there is no such thing as perfection, and that, as Einstein said, “Perfection is the enemy of ‘good enough.’”
Christoph: Right. So, very interesting. And the other thing that was interesting, what you just mentioned is communication. Well, first of all, I grew up in journalism, then I went into corporate communications and marketing.
The term “data scientist” never even crossed my landscape until a couple years ago when I had a short stint in software as a service. And certainly, that is important to learn how to talk to each other. That’s always something to keep top-of-mind.
Jim: I was just going to say people hear the term “data science” or “data scientist” and it’s hazy and it’s very hype-filled. And I’ve got an image in my mind of this spectrum.
On the one side, we have the marketing person, and you can replace “marketing person” with “business decision maker.” Doesn’t matter whether it’s marketing or shipping or manufacturing. It’s a business decision person responsible for PNL and hitting the numbers and doing the work. Next to them is an analyst. And the analyst, so I’ve been dealing with online marketing, so we call them “digital analysts,” used to call them “web analysts” back in the day, and they’re the ones who understand the business problem and figure out how to ask a data question.
Now, the analyst works with a data engineer who is responsible for collecting data, managing, cleaning, getting multiple streams of data together in a pipeline and organizing it—the data plumbing if you will. And then finally, on the far other side is the data scientist, who—imagine in the laboratory in a white coat with test tubes. They’re not building bridges, but they’re inventing new materials. So, we’ve got a data scientist who’s far removed from the business side and a businessperson who is far removed from the high-level PhD mathematics. And a couple of people in between to help bridge that.
Now, in any organization, if you’re a large organization, you might have a team of data scientists and a department of data engineers. And a whole team of analysts dealing with all the different business units. Or you’re a small company and you’re looking for that one unicorn that can do all of it. So, huge company, that’s expensive. Small company, not enough – there’s nobody who knows enough stuff to be able to do that all at one desk. So, it is this interesting challenge of if you are big enough, you can have the team. If you’re a small or medium sized company, then the recommendation is, instead of trying to build this stuff, you’ll buy it.
And you can either buy it from—well, if you’re using Salesforce, they have a machine learning system built in called Einstein that can help you. Or you can go to startups that are creating brand new platforms that are based on machine learning and massive data management. And you don’t have to design and build your own and hire data scientists, you can take advantage of the tools that are being brought out as we speak.
Christoph: And what’s interesting to me, just flashed before my eyes here, when you were talking about the team, I was actually a conferencing partner with OmniUpdate in California, where you are. And somebody asked—I was talking about, how do you instill a data-driven culture? Create content based on performance. And somebody said to me, “Who on the team should play the role of digital analyst?”
I said, “Well, let me rephrase that question with a different position. Who on your team should play the role of writer?” And they said, “Well, the writer.” And I said, “Well, same answer.”
Jim: Right. Exactly.
Christoph: So, how do you bridge those gaps? Things are getting more and more complicated when you don’t have all these roles that you mentioned. And I’m not a big believer in the slash, right? I can’t be the social media manager/web designer/writer/who knows what else.
Jim: Right. Exactly.
Christoph: What kind of tips do you have when it comes to that?
Jim: Well, marketing has become massively complex. When faced with this incredible complexity, we have to go back to first principal, which is, what are you trying to accomplish? And for marketing, we break it into awareness, engagement, sales and keeping people around, loyalty. So, you need to do all of those, you want to do all of those, but you can’t do them all at the same time. What is your primary goal right now?
If you’re a startup, it’s awareness. If you’re rocking and rolling, you want more sales. If the sales are kind of plateauing, we need to engage prospects better. Pick one of those, and then go out and find the tools, the processes and the individuals that can help you, it’s the triad of people, process, technology.
If you focus on solving one problem, then you’ll find that there’s some people that can help you and there are a lot of tools that can do the job. But if you’re thinking in terms of ‘I want to become a data-driven organization next week and I want to become an artificial intelligence capable organization next week,’ it’s just not going to happen. It doesn’t work that way. You solve one problem at a time and over time, you look around and go, ‘Yeah. We’re pretty well data-informed and we’ve got some really interesting tools to help us do it.’
Christoph: Yeah. One day at a time. Quick personal story. This happened this weekend. I was at an Iowa football retirement party and I asked the head coach—and I played there many, many years ago now—and I said, “How is the season looking?”
And he said, “Christoph, it’s not even about the season, it’s about today and tomorrow.” And then Marshall Yanda, who still plays for the Baltimore Ravens, in his 13th season coming up here, and he said, “You know, I have the really good block.” And all the coach said, “Well, good block, Marshall, but your first step was terrible.” You know? “Work on the fundamentals.” Same thing. That’s kind of what I heard you say, right? One day at a time, move forward, try new things and win the championship one day at a time.
Jim: Yeah. You’re not going to become a master online marketing organization by doing some big change management program. You’re going to do it one step at a time.
Christoph: So, how can marketers get started with AI? What’s the next logical step that you would recommend?
Jim: First is education. Figure out what you don’t know and go learn some more. That doesn’t mean that you need to go become a data scientist, but you might want to. That doesn’t mean that you need to get a PhD in statistics. But if you remember, if you have your old college statistics books, go reread them. And there are a huge number of online resources for refreshing yourself.
As a marketer, you need to understand statistics, you need to understand analysis, you need to understand machine learning. But you don’t have to do it. You don’t have to be a programmer in order to advertise on Facebook. And guess what? You’re using machine learning. You’re using Google advertising? You’re using AI tools, even though you don’t know it.
So, the way to get started is to read what you can online, find the case studies, go to the conferences. I would be remiss if I did not mention the Marketing Analytics Summit, coming up in Las Vegas in the middle of June, where we’re gathering analysts together to talk about how we’re doing analysis, what are the common problems and what are the new tools coming up, with some examples and some case studies.
So, keep your ears open and your antennae in the air. There’s a handful of podcasts. I’ll mention two in particular that are my favorite.
And I admit that about 30% of it goes over my head. But just understanding the problems that we’re dealing with at the moment. On the one hand, what great advances we’re making. And then on the other hand, really? We haven’t figured that one out yet; it’s very informative. So, between those—and go to the show notes and follow those links—there’s quite an education to be had.
Christoph: And then, of course, I’ll mention it for you, Artificial Intelligence for Marketing, Jim Sterne’s book. I would highly recommend that as well.
Jim, what do you say to the people, and you probably hear them, too, and I’m seeing people sitting in their office, on their commute at home, wherever they’re listening, and they’re like, ‘I don’t need any of this. My gut feeling has done me well my last 20 years or 30 years or longer.’ How do you respond to them? Have you heard that argument before?
Jim: Absolutely. Sure enough, gut feeling is a powerful tool. And if you team up gut feeling with data, then you have an advantage. I’m reminded of artists who said, “Photoshop? Oh no. I would never use Photoshop. That’s not art, that’s just computers.”
It’s a person with a pencil saying they would never use a pen. Or it’s a person with a pen saying they would never use a typewriter. Or a person with a typewriter saying they would never use word processing. Why would you not want to use a new tool to leverage your gut feeling?
Christoph: Right. Because it’s change, probably. You know what’s funny about you saying Photoshop, I actually had some designers, I don’t remember even what the tool was, there’s a new tool now. And they said, “Oh, we would never use that. We use Photoshop.”
Jim: Yes. Exactly. Fear of the unknown.
Christoph: That’s a never-ending cycle. Fear of the unknown. To wrap us up, Jim, what are some other barriers? Any final thoughts that are worth sharing with marketers as it comes to AI?
Jim: Well, I think people need to be cautious that AI is not a solution. It is a tool. If you have a hammer, it does not make you a carpenter. So, understand what data you have available. Understand how analytics works. Get to understand. Again, reread your statistics book. Because you can’t just say, ‘Oh, this is a problem we can solve with machine learning and I can skip all of those other steps.’
No, actually, you can’t. You are not going to be able to—your new pair of skis will not make you an Olympic skier. You need to learn the fundamentals.
Christoph: The other thing that I found very interesting in the book, I don’t remember the exact example, but it was something about the machine—it was doing what I told it to do, but not what I wanted it to do. So, you still need human intervention and oversight.
Jim: That’s one of the three things that we will always need humans for and shows where AI can’t solve all our problems. As a marketing person or a businessperson, you will always need to do three things.
And if you give it too much information, it is 50/50; it has no confidence at all. What problem are you going to solve? What data should the machine consider? And is the output logical? Common sense. Reasonable. And those are things machines don’t do yet. What problem, what data and the smell test—does the result make sense?
So, when the machine says there’s a correlation between ice cream and drowning, and ‘Oh, my God. Ice cream causes drowning.’ No, that’s not true, it’s the temperature going up that causes both of those things. And they’re correlated, but unrelated. They are not positive. It takes a human to look at that. It takes a human to say, ‘Machine, I want you to create an email campaign that will get me more opens than ever before.’ And the machine will try a bunch of things and figure out that if it sends 1,000 emails to every person every hour, everybody will open one of them just to figure out who to shoot. Not a reasonable answer, doesn’t pass the smell test. But it did what I asked it to, but not what I meant.
Christoph: Yep. Very interesting. Jim, thanks for joining us to talk about AI for marketers. We appreciate it.
Jim: Christoph, thank you for inviting me.
Christoph: And thanks everyone for listening to another episode of Stamats Insights podcasts. Of course, if you’re listening on iTunes or any of the other podcasts channels, always check us out on Stamats.com for more blog posts and additional shows as well. Thanks, everyone.