The time for the question, “Are AI and machine learning here to stay, or are they momentary fads?” has long passed. Now, the answer is a resounding “yes.” Artificial intelligence, and its subset, machine learning, are integral to marketing these days. The real question that marketers need to be asking themselves is: “What is our intent with this technology and how should we be using it?”
AI and machine learning are the buzz of the industry, but many marketers don’t know how to get started. Others have acquired the technology for the sake of having it but fail to develop a usage strategy. This is the equivalent of declaring “I want to make a cake,” without having the recipe or required materials to bake one.
Before any marketer can be truly effective using new technologies to enhance their strategies and tactics, they must understand what they’re using, how to use it, and what their goal is.
If you were baking a cake, you’d need to have a recipe and the correct ingredients to ensure you’re making the cake you envision. In the same way, you need to acquire and understand the tools and data that will allow you to implement AI and machine learning successfully. When you approach marketing in this manner, developing better strategies and tactics becomes a piece of cake.
1. The goal: What are you looking to achieve with your AI and machine learning?
Choosing where to apply machine learning is no different than choosing the right cake for the occasion. Selecting these goals should be driven the marketer’s objectives, the amount and quality of data available and, of course, the types of objectives that the technology excels at meeting. There’s no point in employing artificial intelligence or machine learning just for the sake of it — make sure that this technology will deliver better results than other methods.
A marketer may have the goal of finding more new shoppers during the back-to-school shopping season that are similar to the brand’s most loyal customers. To accomplish this, the marketer must choose the right data, and sufficient data, to train the algorithm about the characteristics of the loyal customers and their back-to-school shopping habits.
If you try to create lookalike audiences of back-to-school shoppers based on a list of Valentine’s day customers, you’re not likely to see great results. The customer’s goals and behavior are very different when shopping for one occasion versus the other. Instead, they’ll want to look at past back-to-school seasons to identify the behaviors that signaled customer engagement and determine what characteristics will help identify other, undiscovered customers.
This approach relies on high-caliber and clean data, and machine learning is essential when you’re dealing with high volumes of real-time data with more data points than a human brain can synthesize. Machine learning can quickly and efficiently identify and leverage the right information, compare the data to the goals the marketer is trying to accomplish, and then provide the results a marketer can take action on.
2. Gathering ingredients = Gathering data
To achieve good results, however, you need both a good recipe (goal) and the proper ingredients (data).
This can feel like a “chicken or the egg” situation. If you’re the marketer looking to improve performance during the back-to-school season, should you consider what data you have available and choose a goal based on that, or should you pick a goal and then review what data is available to help you?
The former is your only viable option. Once you understand what data is — and isn’t — immediately available, you must start with a goal based on what you have on hand to leverage. Meanwhile, you can begin collecting additional data that will help you meet future goals.
To properly leverage machine learning, a marketer’s priority should be procuring usable and clean data. Without clean, reliable applicable intelligence there’s not much a marketer can do -– regardless of what tools they have at their disposal. Valentine’s Day data isn’t going to help find new back-to-school customers.
You can add new goals over time, once you’ve got the necessary data in place, but data should be your starting point.
Marketers who use outdated, poorly managed or unclear data risk hurting their machine learning-driven strategies. At best, this could result in an ineffective strategy. At worst, it could misinform and misdirect a strategy — costing marketers potential revenue, connections and opportunities.
Before even embarking on a project that involves machine learning, you must first evaluate the quality of the available data. If it’s lacking, you may need to clean it up or augment it with third-party data before you can begin to move forward toward your goals.
3. Execute the machine learning-led strategy
There is a common misconception that a machine-learning intelligence product enables marketers to kick their feet up on their desk and let machine learning “do the work.” Machine learning is not a replacement for a marketer; it is a powerful tool that empowers a marketer to be more informed, strategic, and intelligent.
If you don’t take the right planning steps and gather relevant data to analyze, machine learning tools are largely going to be useless.
To go back to the baking metaphor, you know that you can’t produce a high-quality cake, no matter how fancy your oven, if you skimp on the ingredients or if your recipe requires a different cooking method. Think about machine learning as a tool to help you create better, more informed marketing strategies — not as a replacement for them or for you.