Looking back at the past five years in marketing, perhaps the greatest mistake I have repeatedly made was simply coasting, going about each task the way I always have. Instead, I should have constantly sought out new tools that might help.
Machine learning advances in marketing technology have produced an app that can help or assist just about every task in my worklife. In essence, if you have a need, an AI-enabled app probably exists for that. That’s true, from bots that better process image edits to complex neural network algorithms that find similarities in buying behaviors across diverse customer groups.
Every week new players come to market. In fact, each piece of marketing technology these days seems to tout AI in some fashion or form, including machine learning and some associated clever application.
Marketers should see AI (or more appropriately labeled machine learning) as a new feature set that should just be included as part of every digital media tool. Why?
When technology tools remain static and don’t learn they fail to leverage the incredible amounts of data that are available to us. The big data revolution has passed, but are we using are not using data as well as we could. Machine learning helps resolve that. Let’s dig deeper.
Because there is a ridiculous amount of unleveraged data in various states of cleanliness and incredibly difficult marketing automation tools, most marketers execute a multitude of redundant manual tasks to execute their work. This is in spite of the ability to automate them. Worse, marketers are left to interpret the results (e.g. the data) without formal data science training. Or they hire consultants and data analysts to help provide custom reports that probably don’t impact the business.
How many times have marketers missed opportunities because they simply could not possibly comprehend their data?
More than a decade after the web 2.0 revolution, we have created data and missed the opportunity to use it well and mindfully to serve individual customers. Now machine learning offers a chance to successfully build a meaningful and personal customer experience.
The reality of marketing today means that there’s an app for that. It likely has some form of machine learning, and if it doesn’t, developers can help add that machine learning so our systems improve towards business goals. Every marketer should look for machine-learning empowered apps that can aid them in their efforts. This is a great movement to start managing the big data mess.
Cleaning Up the Big Data Mess
Customers and extended communities have given brands great amounts of data. They have opted into cookies, willingly provided personal information, and let brands or their digital media partners store transactional actions and records. These statements assume companies have played by the rules, respecting customer privacy with GDPR and opt-in best practices.
Perhaps companies have melded disparate databases with digital social media and advertising properties and now have an even wider base of customer data. Theoretically, brands should know whenever their customers visit a website or a partner website in various ad networks. Further, they should know what types of content they like.
Yet, most marketing actions are blind, blunted, and aimed at wide swaths of customers with vanilla messaging. Even with customer profiles — the “personalization” tool touted by most marketing automation companies as the answer — brands force customers into preconceived views of how a segment should “behave.” The ultimate goal is conversion with a blind eye towards the other 80–90% of people who will never purchase from a brand unless they have no other recourse.
The end result of the big data mess is incredible amounts of data and crappy spammy marketing. Let’s be clear: Spamming cannot demand loyalty.
Easier solutions may be upon us, When I consider what machine learning can do, I am excited. This enthusiasm has nothing to do with the sentient AI Bot meme that seems to be unfortunately wrapped around this latest technology trend. Nor does it have to do with the latest hype cycle coming out of Silicon Valley, “Everything AI, yay!”
Rather, my enthusiasm centers around algorithms that better identify customer commonalities and trends. Consider most marketers’ predetermined views about how their customer profiles should act based on years of emailing and website behaviors. A basic marketing analysis or report is subject to the marketer or the executive team’s often biased and perhaps wrong views of the customer. Data, clean or biased, is forced to fit these views.
Some algorithms ( K-Neighbor and Hierarchical Clustering to name two) can identify data commonalities that escape fallible human-developed classification schemes. A predictive decision tree algorithm can help test new ways to communicate with customers. Or a social media naive Bayes algorithm can help brands identify online community members that most likely care about similar issues.
Data science can help marketers better process the wide swath of information that’s available and make better decisions! Or work faster in a more productive matter. That’s the real potential. A better approach to marketing not only cleans up data but also addresses human errors and biases.
Machine learning can disarm such prejudices, even the subconscious ones, the biases that every human being brings to the table. Companies like Google actually deploy machine learning tools to not only clean up errant views, but also the inherent biases that data scientists subconsciously include in their algorithms. Perhaps these can be called meta-machines. We will discuss more of the data scientists’ perspective later in the book.
When deployed well, machine-learning approaches in marketing communications have already been proven to be disruptive. We have already seen how algorithms can reshape global markets and even political outcomes. If you need a specific example, just look at Amazon’s incredible dominance in almost every market it enters.
But in the end, we must be wary of the larger AI hype wave. AI bots and the data fed to them are intangibly infallible because they are human. Algorithms suffer from the same fallibility as their creators. The machines are only as good as the people that make them, and the data companies use.
Because AI is fallible it will eventually succumb to overhype and the result will disappoint. That is until the movement arises again under a new buzzword.