Ever wish you had an army of robots at your disposal, the kind that can actually think for you, not just vacuum your floor?
With machine learning, you might just get your wish, as it increasingly enables marketing automation platforms to move beyond merely following your instructions to actively adapting to human behavior and even anticipating it.
This shift has the power to dramatically change our jobs as marketers, says Swede White, media and communications specialist at Wolfram Research, one of the world’s pioneers in computation and computational knowledge. Read on to learn how.
How will the rise of machine learning shape marketing in the next 5 years?
It will replace marketing departments. I’m kidding.
Here’s the thing with artificial intelligence and “the rise of machines.” We still need humans to program them to do what we want them to do. One thing very unique to humans is the concept of goals. Machines don’t have goals. So, first we have to decide, “What do we want the machines to do?” For marketing, and business in general, this, of course, is creating revenue by identifying opportunities, generating leads, and closing sales.
Much of the machine and human-generated data goes into a giant lake and is never organized for analysis due to constraints on human capital to actually do the work. With automated machine learning, we’re starting to get a grasp on creating something like an automated data scientist to wrangle data, reduce dimensions, and get it into some kind of shape so that a human can then query it and gain some insights without having an entire data science department of two dozen people.
We’re even getting to the point with some unsupervised machine learning methods to parse out huge swaths of text to automatically generate things beyond, say, topic models or sentiment analysis. We can actually, through textual network analysis, get to something much more useful, like constructing cognitive schema, from huge amounts of text. So not only do we now know what broader topics people may be talking about, we’re getting into the arena of how they actually think when constructing their language about certain topics.
This still sounds like a lot of work.
It is, but it’s still better than it used to be. Using an example from data science, and, really, marketers are becoming more like data scientists, you start with a ton of unstructured data from all over the place.
Well, it has to be wrangled, dimensions reduced where possible, etc., and that can be a fairly automated process with the right tools. Then you have to see if anything is even there worth exploring with some hypothesis tests. That’s why you need humans.
We can automate some hypothesis tests, but machines aren’t quite smart enough to even know what questions to ask. So, let’s say we want to know whether or not a certain type of content performs better than another, and we get a significant result after our hypothesis test. We can then move forward with some automated tasks that we have defined using machine learning to predict outcomes of the more favorable content, such as whom to target. In my mind, this is crucial to anyone implementing a strategic marketing and/or PR campaign.
So how do we improve this process?
Like I mentioned earlier, we automate tasks across teams, like PR, business development, marketing, and sales, to have everyone’s data communicating.
There are some wonderful products out there geared toward marketers, but I sometimes feel like they fall into the “jack of all trades, master of none” category because these giant platforms try to do so much.
The most comprehensive solution is programming your own solution in an integrated environment. Every organization has its own processes, needs, and goals, and most out of the box software inhibits you from telling the machine exactly what you want it to do. So with something like Wolfram Language, as an example, humans and machines are able to communicate incredibly intuitively to achieve the goals you have defined as successful, such as content strategy.
Worth mentioning is the rise of chat bots. While this is annoying for users on Twitter, it is gaining popularity on Facebook. Also, iOS 11 will have an iMessage functionality to communicate with brands via chat bots. An interesting development is enhanced sentiment analysis to tell the bot how to better respond to a customer/user. Traditionally, sentiment is ranked as -1 (negative), 0 (neutral), 1 (positive). We’re getting a little more granular now. The exciting thing is that chat bots will be able to respond to customers based on perceived emotion to provide them with a better experience. While this is mostly a customer service function, marketing and PR teams most certainly should be paying attention to this.
I want to stress, however, that this is all a means to an end. One of those ends is building relationships and, ultimately, trust. This is important across all of human relations, and business is not immune to the importance of face-to-face interaction, rapport building, and, heaven forbid, having fun in business relations with your clients, partners, and vendors. Computational methods do not replace this. They simply foment and amplify the process.
What is the biggest myth you’ve seen floating around about machine learning and marketing?
First, that somehow Skynet from The Terminator is coming soon.
Seriously, though, the biggest myth is that some kind of out-of-the-box solution is going to instantly give you results and actually uses machine learning or neural nets to provide insight.
It’s unreasonable to expect some kind of dashboard or natural language query to tell you the secrets to selling more products to your customers, gaining new customers, and making billions of dollars, without putting in the planning, work, and, really, programming, to tell the machine what it is exactly you need it to do for what the organization has defined as success. This again is where goals and strategy tie in. For better or worse, machines aren’t smart enough to do that. Yet.