The cynics among us would say that the new year gives people a new start on old habits. And despite great leaps in innovative technology and know-how, when it comes to the world of high tech, some digital businesses seem no different from the masses who fail to achieve their resolutions after several weeks or months of trying.
For the past eight years, companies have dubbed each January "the year of personalization" when it comes to digital marketing. Because businesses have many different types of users who require various types of experiences, "personalization" is deemed part of the solution to the million-dollar question today: how can we predict customers' behavior and thus improve the digital experience? Businesses across every industry invest countless resources in an effort to answer this question, as it will help the organization gain a competitive advantage, and thus increase profits -- but many end up failing repeatedly.
What are they doing wrong?
To answer this question, it's important to realize why digital personalization is important in the first place. The underlying assumption is that if we can identify similar behaviors among groups of customers, this will allow us to predict how other clients will behave. But this is easier said than done. Some believe that, just like buying a new pair of sneakers and some flashy sports clothes in order to join the gym, all you need is new data crunching technology, big data and some data scientist to find meaning in it all, and you’ll be ready to make predictions. The harsh truth is that this is a recipe for disaster.
There's too much data and not enough psychological analysis.
The availability of data is not the problem -- quite the opposite. Today, there is data that relates to user behavior (clicks, page views), social events (likes, shares), item details (category, price) and contextual information (time of day, weather, device) and much more. The problem is that none of it tells you why your customers are (or aren’t) buying your product online. We spend far too much time looking at what customers do, and not enough time trying to understand why they do it.
For example, let’s assume that our data-driven analysis of consumer behavior in a clothing store reveals that the average order value was much higher during daytime compared to evening. The data-based conclusion would be that people tend to spend more money during the day, and thus there’s nothing else to be done. Yet an alternative explanation might be that the customers who came in the morning were welcomed by a friendly, beautiful sales rep who smiled at them as they entered the store. The warm interaction had affected their shopping behavior.
What’s needed is to organize the data in a way that shows a deep understanding of how people think and behave. In other words, consider hiring a psychologist.
Psychology helps us build more accurate interpretations of situations. For example, studies have shown that when we read content from our smartphones, we pick content that’s more emotional than content we might read on our desktops. This can be explained by the fact that we perceive our smartphones as an extension of our bodies -- something that we trust with the most intimate details of our lives -- whereas desktops feel much more separate and distinct, not to mention the fact that they don’t travel with us.
Behavior is more than the sum of its parts.
When it comes to human behavior, data scientists can sometimes be accused of living according to the "black box" assumption, essentially, that human behavior can be understood only by observing external data. As Yuval Noah Harari wrote in his article "Big Data, Google and the End of Free Will," “In its extreme form, proponents of the Dataist worldview perceive the entire universe as a flow of data, see organisms as little more than biochemical algorithms and believe that humanity’s cosmic vocation is to create an all-encompassing data-processing system.”
Such an approach doesn’t work in everyday interpersonal communications -- studies have found that day-to-day interactions are based almost entirely on nonverbal communication. Nor does it work in understanding online behavior. Instead, we have to take into account the underlying drivers of customer decision-making processes. Data analysis should be driven by expert knowledge and psychological theory instead of applying the "let’s just try and see" approach.
One of the data scientists of our biggest retail clients told me he found significant differences in the paths that led users to convert. He found that users who converted weren’t using the website’s filters, while visitors who didn’t convert were using it. He was about to show this finding to his manager and suggest deleting the filtering option until I spotted a fundamental flaw. We know that conversion is a process and it takes few iterations for a visitor to convert. In his analysis he didn’t differentiate between first-time visitors (who needed the filter to find what they were looking for) vs. returning visitors who have already gathered the information needed to make a decision and were ready to convert.
Psychological models can help us better understand the notorious problems of irrationality and the role of emotions, cognitive bias or environmental cues play in our purchasing decisions. People may believe that their purchasing habits are a result of a rational process, but in reality, many of our purchasing decisions are done on impulse, when something triggers our limbic (emotional) system.
Stop talking about conversion rates and start talking about conversion cycles.
We need to start thinking of conversion as a process and not as an action or an event as stated by the conventional wisdom of analysts and data scientists. The path from initial brand exposure to cash register is long and usually it takes few touch points for the user to convert. We need to stop talking about conversion rates, and start talking about conversion cycles.
A “conversion cycle” can be defined as the continuum from product or brand exposure to purchase. This process can involve multiple iterations and may also include transitions between different devices or between the offline and online worlds. Conversion is, after all, the result of an intertwined decision-making process, and the site visit is only a small part of this process. Think of it as an iceberg: what you see above the water is the result of everything hiding beneath the surface. The next challenge is identifying where the customer is in the conversion cycle. When we know this, we can effectively influence his or her behavior.
By segmenting "new" vs. "returning" customers, we can see where the customer is in the cycle and hence offer different recommendations. For first-time visitors, we can provide more information, assistance and popular or bestseller recommendations, while for the returning visitors who know what they want, we can offer a discount to increase the motivation to buy or use urgent massaging such as “limited stock” or “limited edition.”
It is psychological models of customers’ behavior, and not just data, which can help identify intent. It’s possible to translate these patterns of customer behavior into advanced metrics to understand where the visitor is in the conversion cycle. Algorithms could integrate visitor actions, attributes and contexts (such as the type of page or the type of website) in order to determine intent.
In the near future, we believe that businesses will be able to use this knowledge to react in real-time to each individual’s needs, at each unique stage in the conversion cycle. All this will ensure that "personalization" will become a reality, not just another one of those old habits.