Why Email Marketers Need Machine Learning: Q&A with Katrina Knight and David Baker
While many marketers are finding behavioral insights provide essential customer and audience insights that work well when modeled with machine learning, there can be problems, too. David Baker explains in an interview with Katrina Knight of BizReport.
(Originally published in BizReport: 07/28/2017)
Q: How can machine learning best help email marketers?
David: While many have difficulty compiling data from various sources, marketers are getting smarter about identifying and activating the “right” data vs “all” the data. The key here is just processing the many options as fast as you can.
Predicting what content customers might react to best is another benefit machine learning delivers faster and more efficiently. Product recommendations and content recommendations play out well when your goal is increasing engagement and you have many SKUs and you want to see the effect of behavioral data on purchased data. It also plays well with content to help narrow options for personalization.
Q: What about the negatives?
David: While chiefly positive, there are some negatives to consider, without layers to test/optimize you limit the possible variations to typically your strongest, highest converting segments. Lastly, machine learning is impacting how marketers are testing email marketing, using algorithms to optimize campaigns is moving the email marketing space from batch and blast, to a more programmatic view of test/optimization. This is beginning to resemble advertising optimization versus catalogue marketing. It’s an incredibly exciting shift with great potential to change the game.
Q: What drawbacks are there to using AI in email campaigns?
David: The drawbacks to any nonhuman decision making are accuracy and human instinct/intuition. Neither machine learning nor AI is a toy for lazy marketers. You only get out what you invest in them, and they require work as the marketer fueling them must continually tune different models. It also requires careful consideration of a company’s risk profile. With AI, marketers have infinite options, having the confidence to let programs self learn and make decisions on customer experience, is a scary proposition for traditional marketers. The key is to balance predictions and the sheer pace of insights with intuition and in-market discipline. It’s much easier said than done. As valuable as this can be, it can also be a major distraction, if not managed well. Marketers need tools that are proven to work efficiently at their disposal. Without them, it’s too easy to lose sight, slow down and hamper goals.
Q: How do machine learning and adaptive messaging work together?
David: This is a simple concept. Machine Learning is the ability of machine to learn without having to be explicitly programmed. In human words, you don’t need the answers or even the questions to learn, which is tough for marketers who live off of a hypothesis and prove it or disprove it approaches. Adaptive Learning is use of machine and interactive devices to orchestrate human and mediated resources that adapt to each user. They work together best when you have a commitment to long-tail engagement. You have an unpredictable , fluid industry and you have complexities centered on device and place shifting.
At Cordial, we believe the real impact for companies will be in driving value through interactions and allowing machines to help adapt that over time, where the rules may change by lifestage, time of year, or even impact of macroeconomic shifts. Adaptive simply translates to building a value based connection that provides service value, promotional value, education value and entertainment value in sustainable ways.