Over the last decade and a half, as intuition gave way to data-driven approaches to designing marketing strategies, advanced analytic approaches have been used to segment marketable audience bases to enable personalization. Over time, the traditional segmentation techniques (based on recency of purchase, frequency of purchase, and monetary value of the purchase) gave way to more refined segmentation approaches, known as micro-segmentation. Micro-segmentation would leverage more (statistically significant) variables to build a large number of audience segments, each of which was a subject of differentiated personalization strategy (offer personalization, channel personalization, personalized messaging, merchandizing personalization, etc.). Even as marketers saw success using micro-segmentation, ambitious marketers went even further to aspire to develop personalization strategies that were based on each individual customer’s persona (demographic, digital behavioral including both purchase behaviors as well as intent data). Such personalization approaches came to be known by a cool-sounding name called hyper-personalization (a.k.a. one-to-one marketing strategies).
From a puritanical point of view, one-to-one marketing strategies deliver a higher level of personalization, arguably. The abundance and ubiquity of data, the virtualization of data and processing capabilities, and technology hype cycles where theory largely gets promulgated as art-of-the-possible, have the market believing that one-to-one personalization is THE thing. This is evident in our business interactions across the US as curious clients ask us how they can execute scalable one-to-one marketing.
In this post, we present what we explain to our clients and prospects alike why hyper-personalization is not for everyone and why not everyone needs hyper-personalization.
Hyper-personalization works best for businesses that offer ultra-niche experiences to an exclusive set of customers and have access to rich data, ideally exclusive consented data (e.g., ultra-high-net-worth individuals, elite clubs, etc.). Applied to any other business, it runs into one problem or the other. For small and medium businesses, there is the question of data adequacy itself as well as the appetite to invest in data, infrastructure, processing and consumption of data, execution, and fulfillment. For large organizations, the challenge is deciding where to draw the ethical and legal line on data usage.
In our conversations, we commonly come across marketers who think Big Tech, such as Facebook and Amazon, are hyper-personalizing the experiences for their customers, but their personalization models are based on sophisticated micro-segmented lookalikes. Amazon’s famed recommendation engine uses collaborative filtering and resorts to micro-segmentation by analyzing customer purchase history and browsing behavior to suggest relevant products tailored to individual buyer preferences.
Right from the need to significantly enrich individual customer-level data to fulfillment of hyper-personalized experiences, and every data state in between, it is expensive to implement true hyper-personalization. Also, since hyper-personalization relies on defining customer lifetime value (CLTV) instead of lookalikes, it is difficult to predict customers’ tastes and preferences for businesses where frequency of purchase (FOP) businesses is low (E.g., travel, hospitality, cruise). To compensate for limited behavioral data (low FOP), businesses need to compensate with proxies like intent data and expensive third-party data sources, and then spend some more on creating a single version of truth in the data. And yet there are no guarantees, that the downstream analytic models will meet the necessary and sufficient conditions to deliver quality personalization.
Strictly speaking, hyper-personalization models are designed based on an individual's behavioral data. So, they work well on homogenous customer cohorts. Within the defined cohort, it is easy to scale the execution.
Most businesses, especially large businesses, have a heterogeneous customer base and hence many cohorts. These cohorts may undergo changes due to several reasons (new products introduced, data augmentation, product sunsets, etc.). The shifting cohorts pose a problem for hyper-personalization not only because they necessitate constant review and redefinition of the analytics model libraries that power the hyper-personalization engines, but also for the downstream customer experience who may perceive the arbitrary shifts and breaks in experience.
Not only is hyper-personalization expensive to operationalize programmatically, but the data, analytics, and targeting approaches to hyper-personalization may sometimes touch/cross legal and ethical boundaries. Imagine a big retailer using personalization algorithms based on the mining of specific products in the customer's basket! Or companies mining personal Internet browsing data.
In many developed countries, it is outright illegal to mine customers' personal data, but in the countries with ambiguous or lax data privacy laws, unbelievable as it may sound, businesses even mine customers' personal data, (e.g., retail basket data at SKU-level, political affiliations, religious and social affiliations etc.) to build personalization strategies.
Contrary to the perception, most successful brands do not use hyper-personalization but micro-segmentation to personalize experiences for their customers. While Coca-Cola leverages micro-segmentation by targeting health-conscious consumers, Amazon uses lookalike modeling to personalize product recommendations. Spotify curates personalized music playlists and Netflix offers its users tailored content recommendations using similar approaches. Micro-segmentation works well because it is not overtly taxing on the organization’s data assets and yet delivers a phenomenal lift in revenue and customer loyalty metrics.
Research shows that personalization generated by behavioral segmentation has the potential to boost revenues by up to 30%. Research by McKinsey revealed that customer personalization via behavioral segmentation can boost net promoter scores by 20%. The beauty of the approach lies in the fact that while it is only an incremental procedural improvement over the traditional RFM approach, the lift in business outcomes it delivers is disproportionately higher in comparison to RFM.
Micro-segmentation is a highly scalable and cost-efficient approach to enable organizations to personalize experiences for their vast customer base.
We do not blame the marketers and customer experience professionals for their penchant for hyper-personalization. We live in a world of hype; technology hype cycles and overselling sales teams tend to skew the perception of reality. At Evoort Solutions, our goal is to remain honest customer champions, by taking a deep, hard look at the prevalent analytics practices and helping our customers understand what’s best for them.