You may have wondered whether we still need segmentation-based targeting when technology has made personalization-based targeting possible. This piece argues that segmentation will continue to be a key marketing tool because segmentation plays some vital roles that may never be addressed by personalization.
The case for segmentation
Customer segmentation, the categorization of customers into groups based on unique characteristics, has a long history in marketing. It provides the information and insights to match communications, offers, price, and products to each customer group, thereby boosting revenue, profitability, and customer satisfaction. The underlying requirements of a good segmentation scheme are that group membership must be unique (i.e. have unique attributes common to the group and different from other groups), that members respond similarly to marketing stimulus, and that segments are targetable. As such, segmentation could be as simple as grouping customers based on ethnicity, age, gender, or location and as complex as grouping members based on multi-dimensional attributes.
But segmentation creates customer groups because it clusters individuals based on a few unique attributes and overlooks several nuanced differences that occur at the individual member level. The desired end-goal is that businesses have the ability to deliver true one-on-one, personal customer communication and engagement. Judging by the level of success that might be achieved by the corner store owner in a small town, firms realize that customer response and loyalty will benefit more from personalization and individualized intimacy than it would with segmentation-based messages that capture a few attributes of each individual. It’s that desire to achieve the personal intimacy enjoyed by corner stores and small businesses with strong one-on-one knowledge of their customers. Corner store owners know their customers’ interests, their birthday, their wife’s preference, their favorite brand of coffee, average expenditure, how promptly they pays up, and what day of the week they come into the store. Personalization provides the opportunity to deliver that level of intimacy at scale.
Personalization promises more relevance
Personalization in the context of this piece refers to micro-segmentation, the customization of products or delivery of targeted messages and offers based on the unique attributes and behavior of the single customer. In other words, “a segment of one”, and not just referring to the act of “personalizing” by name, such as including a customer’s name in an email or putting their names on receipts. Defined this way, personalization is a very high bar to scale. Although the conceptual benefits of personalization have long been understood and appreciated, personalization has been challenging to implement at scale. As such, group based segmentation has always been a more practical option for marketers. Segmentation creates manageable customer groups that can be practically addressed, targeted, and managed. As a result, segments tend to fall between 3-10 groups so that profiles, descriptions, communications strategies, offers, and cultivation plans can be effectively and distinctively developed for each.
Even in instances where identifying micro-segmentation is possible (such as when lots of data and information exists on each customer), the cost of developing products unique to each segment is high, developing and targeting messages tends to be time consuming, and the incremental benefit over larger clusters quickly becomes unjustifiable.
But today, this limitation of personalization and the “addressability” of segments of one are gradually eroding. Personalization has become a reality for many ecommerce businesses primarily because of digital marketing technologies. Access to huge volumes of customer data, the increasingly lower cost of housing data, the lower cost of computing power to grind these data for insights, and the ability of digital technologies to offer customized content on the fly has enabled firms like Amazon and eBay to deliver personalized recommendations, messages, and offers to customers.
These websites deliver individualized recommendations, customized landing pages, and individualized offers based on robust computer algorithms and dynamic online content delivery systems that review customers’ historical transactions, searches, and traceable navigational paths. These businesses then compare individualized customer behaviors to the historical behaviors of similar purchasers to enhance strong recommendations and personalization. Of all the personalization expectations, however, the distinguishing features are historical transactions, expressed interests, and navigational paths along with a few known attributes tested or mapped back to group correlates. The ultimate recommendation, therefore, comes back to leveraging segmentation analysis that helps inform what one specific customer will appreciate based on what “others” of similar attribute have responded to.
Why segmentation will continue to matter
Given the promise of personalization and the limitations of segmentation, will segmentation continue to matter in the years ahead? While it is an appealing stance to be futuristic and visionary in predicting the demise or irrelevance of segmentation, a look at the current state of targeting and communication suggests that segmentation will be alive and well for a long time to come. Even with the mainstream nature of segmentation and the personalization that occurs with sophisticated online commerce firms, we see signs that firms continue to deliver different forms of communications.
1. Knowledge limitation: We can only know so much about a customer to deliver a unique segment-of-one. Businesses with large customer bases could have a significant transaction and engagement history of all their customers. But attitudes, motivations, and beliefs will be available for only a subset of their customers. While some form of personalization may be possible for customers with very robust and distinguishing attributes, business will continue to reply on segmentation to help form a unique cluster based on available data for all customers.
2. Limited product customizations: Even when you assume that full knowledge of all customers is available, customizing unique products to every customer is impractical for most businesses. While it may be possible for very high-end products for the super rich — unique yachts, homes, and cars — only a finite number of customer products are possible for mass products such as pens, average cars, or fast food. In general, offline products will be limited in the number of products that can be credibly and efficiently delivered by personalization. Personalization will, therefore, continue to be a challenge for non-online firms or those with limited information on each customer. As a result, products will continue to be produced based on the needs of large groups (segments) that provide insights on the common preferences of large audiences rather than individuals.
3. Marginal efficiency: Even when businesses have full knowledge and can feasibly provide customized products for their individual customers, the marginal cost of such customizations will not always be profitable for many businesses. The scale required to market or create these products will not be incrementally profitable compared to the development, marketing, or retention of a segment-based product for these customers. Many customers will end up sacrificing the unique “for me only” products for the more affordable “for us” products.
4. Tiered communications: Even in the few instances where personalization is feasible, messages and offers that apply to all customers, block segments of customers, and to individuals still need to be communicated. For Amazon, for instance, the announcement of the Kindle was a customer-wide announcement regardless of segmentation or personalization. On the other hand, the announcement of a new Stephen King title will go to a broad segment with similar interests in the author or the genre. In short, personalization will not replace multi-dimensional business communication, but rather add another powerful tool.
5. Finite number of relevant attributes: There is a practical limitation to personalization. Unless a business has a small number of customers, the ability to truly personalize is limited. That’s because the number of relevant attributes (i.e. those that impact customer preference or impact how they respond to a business communication) that could be combined to form a unique segment-of-one is finite. Birthdays provide a good analogy: everyone’s birthday is unique until you get to 365, after which there is guaranteed to be at least two people with the same birth date. So, after about 36,500 customers, you are less likely to have individuals with unique no-one-else-looks-like-me attributes that matter to your business. In other words, once you have more customers than the permutations of the variables that are relevant to your business, you can no longer have a segment of one as there will be two or more customers with similar unique preferences.
6.Computational inefficiency: There is a theoretical limit to computation. No computation will be able to create unique differentiated betas that span all the unique attributes of individuals. Models such as regression analysis require that variables must be less than the number of individuals. In addition to this, computational parsimony will identify only variables that are significantly different among groups. As a result, models tend to have less than 10 betas. Tree and classification models will work better for micro segmentation, but visualizing this becomes impossible once numbers reach tens of thousands. Personalization, therefore, tends to be based on user activity such as pathing and triggered activities, rather than individualization identified from rigorous computation. Look-alike models are not really personalization, but segmentation schemes that try to classify new individuals into pre-defined segmentation schemes.
Segmentation is here to stay
Just as county and city based communications will continue to be relevant despite presence of addresseable households in a state, segmentation-based marketing will continue to be relevant despite the feasibility of personalization. Segmentation will continue to play a vital role in marketing because it provides insights on customer understanding, product development, and actionable targeting. In fact, it’s arguable whether personalization as defined by a segment-of-one will ever be a possibility for more than a few businesses that adhere to the limitations noted above. Even for businesses to be able to offer personalization, insights for group-based product development, offers, and communications will be a relevant contribution to overall marketing performance.
Digitaliy 
Discussion
No comments yet.