Retail Marketers Put Data in the Driver's Seat for Personalization and Precision
We’re saying it: personalization is queen among today’s consumers. Just look at Spotify, which algorithmically delivers a unique 30-song playlist to each of its tens of millions of users every week. Meanwhile, Google Home's Assistant and Amazon.com's Alexa use voice recognition technology to tailor their responses to each member of a household. Nowhere is this truer than in retail, where personalized promotional emails have a 29 percent higher unique open rate and a 41 percent higher unique clickthrough rate than boilerplate messages.
Unfortunately, despite their best efforts, many brands continue to struggle with delivering the kind of personalized experiences consumers crave. “Nearly 90 percent of organizations say they're focused on personalizing customer experiences, yet only 40 percent of shoppers say that information they get from retailers is relevant to their tastes and interests,” Forrester Principal Analyst Brendan Witcher told The Wall Street Journal.
This disconnect stems in large part from the outdated way in which many brands approach segmentation. By pivoting from a traditional audience segmentation approach based primarily on demographic factors to one based on data-informed customer personas, brands gain the ability to place the right message in front of the right people at the right time — the (hor)crux of effective marketing.
Traditional Segmentation Vs. Behavioral Segmentation
The goal of audience segmentation is to link a set of shared customer characteristics to a series of specific actions. Until recently, most brands segmented their audiences along demographic (e.g., age, gender, ethnicity, location, income) and/or psychographic (e.g., interests, opinions, attitudes, values) lines.
Under the right circumstances, this approach can be quite effective. Customers who make over $200,000 a year are almost certainly going to have a greater interest in high fashion at a high price point than customers who make under $50,000 a year. That said, a luxury designer would be wasting precious time and resources by reaching out to both the “glam Anna Wintours” and the “hoodie-loving Mark Zuckerbergs” of the world with a static, blanket marketing strategy.
This example reveals the limitations of a traditional approach to segmentation: demographic and psychographic characteristics don’t always correlate with a specific set of actions. Indeed, the only way for brands to segment their audiences in a way that enables accurate, consistent personalization is to build customer segments — or “personas” — not around age or attitudes, but around past actions and behaviors.
Ever heard the saying “you are what you eat”? While we haven’t turned into walking hot Cheetos and Takis yet, this phrase does provide a concise summary of why behavior-based segmentation is more effective than demographic- or psychographic-based segmentation. If our actions — eating, or for our purposes, shopping — constitute our truest selves, our past actions provide the most accurate indication of how we will behave in the future.
Of course, in the context of a large retail operation, it’s next to impossible to stitch together an individual’s, let alone an entire audience’s, historical transactional data in a way that produces any sort of meaningful insights. This is where customer segmentation and data-driven personas from artificial intelligence (AI) and machine learning based on statistical patterns within transactional data come in.
Unearthing Personas Via Machine Learning
Networked point-of-sale systems and the rapid rise of e-commerce have given brands access to all the historical transactional data they could ever want. The challenge is making sense of it all. One way to tackle this challenge is by feeding massive volumes of item-level transactional data into a set of machine learning algorithms based on what’s known as the Dirichlet Multinomial Mixture Model.
The algorithms are used to sort through a brand’s transaction data and pinpoint connections between various ordering patterns: if a customer has purchased a series of items, x, there’s a y percent probability that he or she will also take action z. Once these probability distributions are established, the algorithms look for bigger picture purchasing trends to serve as the foundation of customer personas. As with any kind of probabilistic modeling, these personas are a simplification of the nuanced behavior of real customers, but they have proven to be remarkably adept at forecasting future purchasing behavior.
Say, for example, that our algorithms discover five persona templates — fashionistas, hipsters, outdoorsmen, nightlifers, and aspirational shoppers — that, taken together, account for most of the seemingly random variation in a retailer’s data. Based on a customer’s order history, they're assigned to the persona with the highest probabilistic match. If a customer has a probability above 50 percent, they're categorized as “true” to that persona; if they have a probability below 50 percent, they're categorized as “leaning” toward the persona to which they have the greatest affinity.
By segmenting its customers according to these personas, a retailer will be able to deliver personalized messaging with unparalleled precision. For instance, if a customer — let’s call him Harry — registers as a “true” outdoorsmen, the retailer will know to target him with promotions related to flannel outerwear as opposed to dobby dress shirts.
While the ability to create more personalized messaging, especially in channels like email, is an obvious benefit of segmenting audiences by personas, it’s certainly not the only one.
Brands can use clustering algorithms to build lookalike audiences based on their most valuable or strategic personas, which can then be leveraged over digital acquisition channels like Facebook. With the help of a web developer, brands can even curate a customer’s on-site experience by serving them content that aligns with their specific persona’s preferences.
Ultimately, as Forbes contributor Shep Hyken points out, “There's a $2.95 trillion prize for companies that integrate a smart digital strategy to personalize customers’ experiences.” For retailers, persona-based segmentation all but guarantees a share of that prize — not through magic, but through the power of data.
Corey Pierson is co-founder and CEO of Custora, a provider of advanced customer analytics for retail.
Corey Pierson is Co-Founder and CEO of Custora, the leader of advanced customer analytics for retail. Prior to Custora, Corey worked at IDEO and was Co-Founder of foodtrux.com. Corey received his BSE in Computer Science Engineering from the University of Pennsylvania and his MBA from the Wharton School at the University of Pennsylvania.