Artificial Intelligence: Trust the Machine to Master Marketing Automation

magnus

Magnus Unemyr, Marketing Automation Expert, helps companies install and set up website-based marketing automation systems. He recommends using site lead magnets (calls to action, buttons, banner ads), caging some content behind landing pages with registration forms, offering incentives to register information e.g., a PDF that can only be downloaded in exchange for contact information, and adding nurturing sequences with follow-up emails. (He uses Tripwire.) Each piece of content should drive the customer or potential customer one step closer to making the purchase.

In this interview, Magnus clarifies the difference between today’s narrow AI software and strong AI. Narrow AI learns from data and self optimizes over time by iteratively improving its original function. Narrow AI in email application might learn the optimal time to deliver an email to  an individual customer to increase the likelihood of that email being opened. Strong AI has the capability to learn things outside of its originally targeted function . . . to reason on its own, to start to get feelings . . . which is still the stuff of science fiction.

Traditional marketing automation systems; e.g., HubSpot, ActiveCampaign, or Act-On, are fairly basic in autonomous decision-making. Magnus sees a natural progression from using AI-powered algorithms/marketing automation systems to automatically harvest data, eventually moving toward more complex functions – developing “insights,” and, eventually, making autonomous decisions about and triggering individualized marketing outreach initiatives without human intervention.

This future AI will be able to do highly personalized outreach – to send the right content to the right person at the right time, in the right channel, and at the right frequency – improving the customer experience and minimizing the “spamminess” of cookie-cutter automatic responses.

However, when deploying new marketing automation systems, companies need to budget for content production, and each piece of content should send the reader or viewer one step further toward the purchase. Also, major marketing automation tools are not standalone solutions – they need to be integrated with additional specialized smaller marketing automation systems – webinar platforms, proprietary databases, etc.

AI can fail. For instance, AI algorithms are trained by historical data. Without enough historical data, any AI system will not be able to make accurate decisions. If the data is skewed or flawed, the AI algorithm will produce flawed predictions or behavior. The software supplier needs to be able to prove that the software is behaving and producing accurate results.

Magnus describes his book, Data-Driven Marketing with Artificial Intelligence, as the definitive guide to understanding and using AI in marketing. He has also written: Mastering Online Marketing, Internet of Things, Turn your Knowledge and Skills into a Profitable Online Business, and eBooks and Beyond. Links are to Amazon.

Every marketing automation system vendor provides video courses to teach users about their product, but invariably fails to discuss the features that should have been included . . . but weren’t. Magnus created a detailed marketing automation course (about 70 videos) that explains concepts, strategies, use, and what comprises a good marketing automation system – without covering proprietary instructional material. The credit-card accessed course is available on his website at: http://unemyr.com.

Magnus can be reach on LinkedIn or on his website at: http://unemyr.com, where you can find his blog, his video episodes, and his training courses.

ROB: Welcome to the Marketing Agency Leadership Podcast. I’m your host, Rob Kischuk, and I am excited to be joined today by Magnus Unemyr, Marketing Automation Expert, speaker, and author of several books, including Data-Driven Marketing with Artificial Intelligence. Magnus is based in Sweden. Welcome to the podcast, Magnus.

MAGNUS: Thank you very much for inviting me.

ROB: Very glad to have you here. Why don’t you start off by telling us a little bit about your fascination with marketing automation and what has drawn you even deeper into melding that with AI and then sharing that expertise out?

MAGNUS: Sure. I’ve spent over 20 years in the international software industry doing mostly marketing for two companies in the software industry. I have a fascination with both marketing and internet marketing on one side and about software on the other side.

I think that marketing automation is the perfect combination of the two because it is programming business logic that explains how software robots should generate touchpoints automatically and nurture leads automatically towards the purchase. But traditional marketing automation systems like HubSpot or ActiveCampaign or Act-On or any of those software tools have traditionally been fairly basic in terms of autonomous decision-making.

The next natural progression, I would like to say, is to use AI powered algorithms or AI powered marketing automation systems that can harvest data for insights automatically and then make autonomous decisions that decide which kind of marketing outreach should be triggered toward individual leads and customers without human intervention.

This is why I have been increasingly interested in reading more and more about how AI and related technologies can be used to improve automated marketing.

ROB: Really interesting. A lot of times, some of these more advanced techniques and tactics start at the top end of the market and work their way down. Originally even marketing automation itself was fairly inaccessible for all but the smartest.

I think some people are still struggling to adopt the best practices of marketing automation. When it comes to integrating AI, where are we at in the market? What types of organizations are finding that they are capable, and maybe some examples of how it comes into play in a specific example?

MAGNUS: First of all, I would like to say that we don’t really have artificial intelligence or AI. What we have is more like machine learning and predictive analytics. Let’s still call it AI. Within AI, we have two types.

One is narrow AI, which is basically what we have now. That is a software solution that can learn by themselves and self-optimize to do a better job over time, but only within the initial task they were initially trained to do. For example, an email software can learn at what times of the day it is best to send emails to a particular lead, and it can do a better and better job of that using narrow AI.

Strong AI, on the other hand, which we don’t have, is software solutions that can learn things on its own – things that it hasn’t been preprogrammed to do – reason on its own, perhaps start to get feelings, and eventually maybe take over the world. We don’t have that currently.

So, what we have is really machine learning or narrow AI, if you like. That is basically software that can be trained from data, and over time adapt its behavior as the data patterns are changing.

If you look at the use cases for AI or machine learning in marketing, there are hundreds of different things you can use it for. I think I outlined 100 or 200 different things in my latest book. A very simple example is to have email software automatically adapt to send time for each individual lead. Instead of sending a mail blast to 100,000 people at the same time, send the mail blast one email at a time, perhaps at 100,000 different times, to make sure that each individual recipient receives the email at the time that has the highest chance of getting that person to open the email.

But, if you look for example at Gmail or even some of the Apple solutions, Apple Watch, if you get an incoming email or an incoming message, the email client or the Apple Watch proposes a suitable response. You can just click one button to send that response away. That is kind of AI as well, or machine learning, whereby you get automated responses presented.

So you don’t really need a very expensive software solution to start to use machine learning if you use for example Gmail. You may use it already every day.

ROB: Right. Gmail is really good, almost strangely good at times, at suggesting how you could respond. It’s an interesting danger, I think, that we all start to sound the same in using the responses in the current state of AI. Do you think that someone like Google would continue to adapt those responses more towards your personal tone?

MAGNUS: Absolutely. You have already chat robots or automated email robots that can detect which kind of emotional mood someone is in. If we receive an incoming email from a potential customer, by analyzing the text of the incoming email, we can get a measure of how strongly positive or how strongly negative that person appears to be. This is called sentiment analysis. We can use that to craft, for example, different email copy in our response based on how happy or unhappy that person seems to be at this particular time.

ROB: It sounds like you’ve given a few examples where people who are running products should be thinking about how they can be using different forms of AI or machine learning to make their product better.

When it comes to individual marketing practitioners, do you think they will primarily be using AI and machine learning as it’s presented in tools that they use, or will there be specific cases that they should also be more technical practitioners who need to be thinking about how to execute on the project themselves?

MAGNUS: I think that the vast majority of marketers or people in general will use AI and they may not even be aware of it. AI, or machine learning, for the most part will be hidden behind the scenes or under the bonnet of whatever software they are using. It will just be there and improve the behavior of the software, and you don’t need to know anything about AI or how it works. You don’t even have to be aware that machine learning is powering that software to make it operate in a more efficient way.

ROB: Perhaps one of your core beliefs, then, would be that marketing automation is interesting and useful and effective, and it’s only going to get more interesting and useful and effective as we bring in some of these additional machine learning opportunities?

MAGNUS: I think what we’ll see is that we will get rid of some of the spammy mass marketing approaches whereby everyone gets the same email copy, for example, and we will, with AI, be able to do highly personalized outreach – for example, to send the right content to the right person at the right time, in the right channel and at the right frequency.

With AI, we can make marketing automation much, much more personalized and improve the customer experience and become far less spammy.

ROB: How do you suggest people start to think about getting comfortable with that? Is it something that marketing automation practitioners are going to have to get used to, trusting the machine? Or do you think there’s also a burden of proof back on the tools providers and the vendors to demonstrate that they’re delivering value by asking people to trust the machine? What’s the metrics?

MAGNUS: I think that you need to put some burden on the suppliers of the software solutions, and they will have to test their solutions, obviously. But I think it is up to them to prove that the solution not only improves the results, but also doesn’t misbehave and start to send 100,000 emails to some person by mistake because some data pattern indicates something.

Any kind of machine learning solution or AI solution requires a lot of data because the AI algorithms are trained by historical data. If you don’t have enough historical data, then you either can’t create an AI solution that makes accurate predictions, or the predictions are there but not accurate enough. So for sure, you need a lot of data. I think it is really up to the software supplier to show that the solution is actually working and not misbehaving.

A good example of this would be Microsoft’s Tay. I think it was in 2016, so now 3 years ago, Microsoft deployed a chat robot on Twitter that learned the lingo of people who were chatting with it. A number of right-wing extremists started to chat with the Microsoft chatbot, and it learned the lingo of those right-wing extremists. The Twitter robot itself started to spread really extreme political views, and Microsoft had to shut it down within 16 hours or something like that.

So, if you train algorithms from data, if the data is skewed or flawed, then the predictions or the behavior of the AI algorithm will be flawed as well.

ROB: Sure. I saw it mentioned by somebody yesterday. They said that these algorithms are trained by data. The data often comes from people, and people have biases and tendencies. So, of course, we are going to create algorithms with biases and tendencies. Do you have any thoughts around how to present better data into these algorithms? How do we get algorithms that achieve as our higher selves instead of who we are?

MAGNUS: That is the job of data scientists. The data scientists are basically statisticians. To my knowledge – I have some knowledge in software science as well – it is the data scientists that create the AI or machine learning algorithms. They have to prove that the algorithms they come up with really are working and are not biased. In certain situations and in certain countries, it may actually be illegal to have AI predictions that are skewed – for example, for political views or religious views or things like that. You will have to be a little bit careful depending on what kind of autonomous decisions you are making based on what type of data.

ROB: Got it. Using some of these algorithms, some of this data in the tools that you have is going to be one of those opportunities that marketers have to set themselves apart. When it comes to doing marketing automation well today, what are some other key behaviors or practices that you believe someone needs to be doing to be effective with marketing automation versus ineffective?

MAGNUS: There are two specific points I would like to mention. The first one is that content marketing works without marketing automation, but not so efficiently. Marketing automation, on the other hand, doesn’t work at all without content. So, marketing automation is just a smart way of delivering the right content to the right person at the right time.

When you deploy a marketing automation system in a new company, it is quite often misunderstood how much content production goes through that. You really need to budget also for the content production, because the marketing automation system just delivers the content, and if you don’t have that content the marketing automation system can’t really do that much.

Secondly, I would like to say that marketing automation may be important, but my thinking is that marketing integration is even better. Marketing integration, to me, is when you combine the marketing automation system – like HubSpot or ActiveCampaign or something like that – with additional specialized smaller marketing automation systems. For example, with your marketing platforms, webinar platforms, integration with proprietary databases and so on.

Then you can create a coherent, well-integrated marketing automation concept that does far more than just have one marketing automation system as the center point.

ROB: You mentioned the difficulty of creating the content. Within that, even, I think there are a couple of problems there. There’s actually doing the work and having the discipline to produce the amount of content that’s needed.

I think there’s also the question of what content will be effective. How do you suggest people think about creating content that isn’t just marking time and tagging the base and producing content, but actually producing content that potential customers will care about, that reinforces the brand positioning, and really content that delights?

MAGNUS: Content that delights is not really my expertise because that is more copywriting. But I have some strong thoughts about content strategy, which is something different. My expertise is not how to write actual sentences or paragraphs that trigger emotions and stuff with potential customers, but in my mind, every piece of content you create should drive the customer or potential customer one step further down the customer journey towards the purchase.

If you, for example, create a blog post, then that should offer or promote, for example, a whitepaper PDF download or a webinar. Whatever whitepaper PDF you offer, then that PDF should offer the next PDF along the customer journey, one step further closer to the purchase. It’s quite important that you have an overarching content strategy whereby the content fits each buyer persona in each stage of the customer journey, but also that every piece of content sends the reader or viewer one step further along the way towards the purchase.

On top of that, of course, the piece of content then needs to be engaging, have nice graphics, have selling copy, and stuff like that.

ROB: For sure. You also mentioned the value of the marketing integration with webinar and video platforms and that sort of thing. Are there any particular platforms that are out there that you’re seeing are both effective in doing videos or webinars and are also very agreeable to integration? I’d imagine quite often even information on how far somebody got into a webinar, if they completed it – things like that are helpful in deciding the content that should be provided to them next.

MAGNUS: There are several video marketing platforms. I usually use Wistia. With Wistia, you can actually stop the video playing in the middle of a movie. For example, you put a video on your website with a triangle play button on top of it, and whenever someone clicks the play button, the video starts to play for 15 or 20 or 25 seconds.

Then the video actually stops, and you can’t continue watching the rest of the video until you have submitted your contact information in the registration form that sits in the paused video frame. So the video itself becomes a lead generation asset. The video then has an integration to the marketing automation system such that the contact information that video watchers enter to unlock the rest of the video is automatically fed into the marketing automation system contact database.

In terms of webinars, for example, I use EverWebinar for on-demand, evergreen webinars that are self-playing at any point in time throughout the day or week. For live, manually hosted webinars, I usually use GoToWebinar or GoToMeeting. Both of them integrate quite nicely with most marketing automation systems.

So whenever someone enrolls or signs up for the webinar, then we can automatically send an email sequence that reminds them to come to the webinar, perhaps warms them up a little bit with some content, and then after the webinar it can detect whether or not each person actually attended the webinar or not. Then we can send different email follow-up sequences to those who registered for the webinar and actually attended and a different follow-up sequence to the people who registered but did not attend.

For example, to the ones who didn’t attend, we can offer them a replay. To the people who actually attended the webinar, we can send them to an online booking calendar and ask them to schedule a consultation call with a salesperson, for example.

ROB: Interesting. Those are some good tips there. You also mentioned – I think this is generally true on video, but perhaps the pressure is even higher with the strategy you just talked about with that 15-second delay to ask them to fill out the form.

It’s critical in any online video channel to capture somebody’s attention very quickly, but even more so when you’re going to ask them for their contact information after 15 seconds. How do you think about creating the level of interest and engagement in those first 15 seconds to entice the prospective customer to continue?

MAGNUS: First of all, I would like to say that this type of concept probably does not work very well for the corporate style product, presentation, sales video – which is just 30 seconds anyway. This strategy works better with educational or informative videos that have specific value, and someone watches it because they want to learn something from it. That is the first consideration.

You need to create the hook. For example, you can start by saying that at the end of the video, we will offer you this free guide or free eBook or free thing that is of value, and that will keep people watching the complete video, hopefully.

ROB: You certainly walk the talk with what you’re doing for yourself as a practitioner. Tell us a little bit about your online course in marketing automation, how you came to have the conviction that that needed to be out there and would be effective, and what’s in there as well.

MAGNUS: I launched the new online video course on marketing automation a couple of months ago. The reasoning behind that is that first of all, I see there are so many companies and marketers who either don’t know what marketing automation is, or they know what it is but are afraid of it.

I felt there was a need to have some kind of product, independent marketing automation training. Every marketing automation system vendor has video courses that teach you how to use their specific solution, but it doesn’t tell you what features they don’t have in the software that should have been in there. So I felt there was a need for a fairly detailed marketing automation course that explains the concepts, the strategies, how to use it, what you should expect from a good marketing automation system, without going into the nitty-gritty details on how you operate in a particular software product.

This video course is about 70 videos. It’s obviously offered from my website, and you can buy access using your credit card and in a minute or two you’re up and running. You can take the course and become a marketing automation expert.

ROB: Very cool. I now want to be a marketing automation expert, along with 25 other things I have to do in my day. How do you suggest someone who is a marketer but is pulled in many directions – what are some of the baby steps they should be taking to be just two steps better than they are right now on marketing automation, or lack thereof?

MAGNUS: I think you should install a marketing automation system behind the scenes of your website. Make sure there are a couple of lead magnets that are offered on the website using calls to action, buttons or banner ads. Then cage some content behind landing pages with registration forms. So you offer an incentive to register information by offering some PDF that can only be downloaded if they give you their contact information.

But then you need to have some nurturing sequences with some follow-up emails. You can install Tripwire on the website. For example, a certain page view can trigger the marketing automation system to send three or five nurturing emails. For example, whenever someone visits the pricing page, then 2 days later we start to send free follow-up emails that are suitable for anyone who visited your pricing page.

Anyone who came to your product page related to windsurfing boats or motorcycles, for example, will trigger a marketing automation workflow that sends a few follow-up emails that relate to interests of that specific page, where we install a Tripwire that triggers that nurturing sequence whenever someone comes to that page.

Or for example, if you’re a consultant, whoever comes to your consulting page for the third or fifth time will be offered a consultation call. But whoever comes to the page just once or twice does not get those emails.

ROB: If someone just waves the white flag and realizes that they need help from someone who is not them, are you someone who engages directly with people? Or do you have a network of people you recommend if they realize that they need help?

MAGNUS: Most of my business is me acting as a marketing automation consultant. I help companies install and set up marketing automation logic behind or on their website. That is one step, obviously. It is of course possible to take the course as well, to learn for yourself. But I do offer consulting services to set up marketing automation on a company’s website.

ROB: Very good. What are the next couple of places you think that we will start to see AI and machine learning creep its way into our marketing automation? You’ve mentioned send times, you’ve mentioned a little bit of frequency. Some of that is probably closer and some of that is probably further out. What are the next couple of features you think are going to be working their way in and helping us be better?

MAGNUS: I don’t think I can answer that question, because you will see AI or machine learning everywhere, in every corner of the marketing industry. For example, there will be machine learning solutions to help improve content, so content strategy or SEO optimization.

Conversion rate optimization, where you redesign a landing page to get more signups in the registration form. For example, there are now CRO tools that are machine learning based that know how to redesign a landing page hundreds and hundreds of times. For each iteration of the redesign, it measures what design changes seem to work better than others. When it finds the redesign, generation after generation, eventually, maybe after 50 or 100 or 200 redesign attempts, it comes up with the redesigned landing page that produces perhaps 30% or 40% more registrations.

But the really cool thing, I think, now – in addition to what we talked about before like personalized marketing automation and chatbots – is the capability to harvest data from internet connected machines. There will be something like one trillion internet connected devices by 2025, Internet of Things or IoT devices. All of them will generate oceans of data. If you have time, I could just mention two examples, one from B-to-B and one from B-to-C.

Assume that we are the manufacturer of windmill farms. We can actually put sensors on each individual windmill and measure heat and vibrations. By measuring and analyzing the heat and vibration patterns of a particular windmill, we can detect when that matches or correlates to the heat and vibration patterns of other windmills that broke down 2 weeks later in the history. We can detect now that a particular windmill is likely to break down 2 weeks from now.

So, we can use that insight that is harvested by AI, analytics of machine-generated data, to trigger marketing automation logic to offer the owner of that windmill spare parts or services in advance to prevent the machine from breaking down in the first place. That is marketing campaigns that are automatically triggered by machine-generated data, let’s say in industrial settings.

Another example would be that we have, for example, a bathroom scale that is internet connected. I have had one for 5 years now which is internet connected. By analyzing my weight change patterns and correlating that to the weight change patterns of other people who later gain weight, the bathroom scale can conclude now that I am quite likely to gain 10 kilos or 10 pounds in 2 months from now. Therefore, if I’m a member of a weight watching club, for example, they can start to send me healthier recipes or tougher training programs in advance to prevent that weight change from happening in the first place.

ROB: Excellent. Very interesting examples. Magnus, when people want to find you, I imagine a place they should find you – and we’ll put this in the show notes – is unemyr.com. That may be tricky for people to understand from my pronunciation, but http://unemyr.com. Where else should they find you and learn a little bit more?

MAGNUS: They can find me on LinkedIn. Just search for Magnus Unemyr on LinkedIn and you will find my page. As you mentioned, unemyr.com is the place for my blog, for my video episodes. You will find my books and my training courses there as well.

ROB: Very good. Thank you for sharing your deep knowledge on this topic, Magnus. It has been a pleasure. And congratulations on everything you have known, learned, published, and spoken about.

MAGNUS: Thank you for inviting me.

ROB: All right. Take care.

Thank you for listening. The Marketing Agency Leadership Podcast is presented by Converge. Converge helps digital marketing agencies and brands automate their reporting so they can be more profitable, accurate, and responsive. To learn more about how Converge can automate your marketing reporting, email info@convergehq.com, or visit us on the web at convergehq.com.

Leave a Reply

Your email address will not be published. Required fields are marked *