Marketers can harness the full potential of algorithms and reduce the risk of drawbacks when they supervise use, monitor performance, and inject humanity.
While marketing algorithms shoulder the burden for many important tasks for humans, the need for human management and supervision still makes people an essential part of the equation long after they polish off an algorithm’s code.
We recently covered some of the most important uses of algorithms in marketing while generally defining algorithms as a whole. In sum, algorithms are a set of “if-then” rules, leveraging customer-specific knowledge to assist marketers in automating repetitive or complex decision-making tasks, making operations more efficient and less tedious. In the meantime, marketers can focus their efforts on higher value-added tasks – such as collecting and structuring the output of data insights – allowing them to actually connect the dots and apply this knowledge to their creative story telling and long-term brand strategies.
However, we only partially touched on the continuing fusion of humans working alongside algorithms to accomplish marketing goals through a combined effort. Put simply, humans and algorithms work better together. Algorithms are tools, not taskmasters. Recognizing their strengths while supervising their potential weaknesses gives marketers more advantages than ever before.
When people remain closely involved in the management of projects involving algorithms and oversee their functioning, they open a path for accomplishing more at scale than either could do on their own. Secondarily, they must continue to insert humanity into algorithms, as technology cannot easily perceive emotion and context the way people inherently do. To better understand why humans are needed to ensure algorithms function properly, consider the following three critical management components.
Transparency Ensures Marketing Algorithms Function as Promised
Many programmers bestow algorithms with machine learning capabilities, where algorithms can optimize themselves in light of performance-based criteria and the discovery of unexpected correlations. These learning-based rule changes can create “eureka” moments for brands, discovering connections between marketing practices and successful conversions that would not otherwise be drawn.
However, machine learning capabilities can also lead algorithms to draw questionable conclusions unless human supervision is there to monitor them. All automated rule optimizations must therefore be transparent, discrete, and verifiable so that gradual rule changes eventually do not create risks.
Automated rule changes can be governed by a set threshold, for instance, to ensure that dramatic changes do not happen abruptly. Many algorithms already perform this sort of action by flagging instances where they are unable to draw a clear conclusion. For example, one online travel agency uses an algorithm to eliminate duplicate entries when displaying hotel results, but “low confidence” records are flagged for human review. In the same way, major proposed changes to marketing algorithms can alert supervisors and request permission before crossing the established threshold.
Eliminating Algorithm Bias Breaks Down Barriers Through Technology
Certain algorithms pose another common challenge when they reinforce existing societal echo chambers and biases. To provide a harmless example, social media newsfeeds try to display the most relevant posts based on the friends you seem to interact with most. But over time, certain friends can drop off your newsfeed altogether, even if you have a deeper connection with them than a digital acquaintance whose posts you see every day.
Sometimes, these outcomes do not even result from inherent, obvious design flaws. One recent study noted that an ad for job opportunities in science and technology showed more frequently to men than women – not because women were less likely to click on the ad or less likely to be qualified for a science and technology job, but rather because young women are a more sought-after demographic, making them more expensive to target.
By noting the potential for something as innocent as a cost-control parameter to yield bias, marketers can root out inequity rather than reinforce it. These revelations can come from substantial audits of algorithm results, as the study enacted. The Harvard Business Review also recommends studying correlations established by algorithms and spot checking results, evaluating algorithm performance as you would a scientific hypothesis or a product before launch.
Keep Marketing Campaigns Focused on Humans, Not Machines
An ecosystem of algorithms speaking to algorithms has helped the internet become the rich cultural tapestry that we enjoy today, but marketers may want to recognize and stymie the temptation to position their campaigns toward algorithms rather than humans.
For instance, a news organization now uses algorithms to craft certain short news articles heavy on factual details and light on analysis. Since social media algorithms choose news stories people are likely to find interesting, developers run the risk of crafting news algorithms that social media algorithms are more likely to display, but that people may not necessarily get value from.
Similarly, an algorithm that chooses to show someone paid social media ads based solely on that user’s “likes” may fail to predict the potential interests of that user, and in turn, lose out on discovering new audiences. In this way, newsfeed algorithms can fatigue their users or cause them to tune out by displaying repeated ads without truly capturing the broad landscape of someone’s prospective interests.
In the words of Jonathan Nelson, CEO of Omnicom Digital, “The silos between technology and creativity need to be broken, once and for all. There is a difference between an algorithm and an insight. There’s a difference between a touch point and a true connection. There’s a vast divide between big data and a big idea. Old-fashioned intuition is still very much relevant in marketing. Machines need analysts and creatives to prioritize the information and figure out what people really care about.”
To ensure that advertising and marketing can retain the richness it currently exhibits without unintended consequences, marketers will want to be continually open to serendipity and discovery in the campaign creation process rather than relying on suggestions from algorithms alone. Brands can consider their audience from a 360° perspective by testing new segmenting variables, audience preferences, and correlations instead of relying completely on a flattened view of demographics that “fits the model” conveniently for algorithms.
So, even though humans are not part of algorithms’ code, they have become inextricably woven into their use. By monitoring the results of algorithms, taking proactive steps to ensure that common challenges do not grow into larger problems, and appyling creativity and human ingenuity, brands can harness the full potential of algorithms to truly revolutionize our future.