What Eastern Mysticism can teach Marketers about Segmentation
“How many segments should we have?”
“Do you define all your segments in advance?”
“I’m extremely familiar with a certain specific type of customer. What segment are they in?”
The above quotes are paraphrased questions from three distinct C-Level executives representing organizations with over $91 B in market capitalization combined. These questions and similar ones are frequently posed by senior marketing leaders to myself as well as MarTech aficionados such as Alex Uher (a strategic marketing leader, who has worked with global brands such as L’Oréal and Coach).
Mr. Uher and I observed that it is common for some CMOs and marketing VPs to attempt to rationalize and summarize marketing segmentation concepts in a linear and rigid fashion. While we sympathize with reasons doing so (e.g. transparency), the result is often an oversimplified approach that can limit marketer and IT collaborative potential. A failure to recognize and embrace the complex reality could mean missed opportunities to improve business critical metrics.
In Capra’s “The Tao of Physics: An Exploration of the Parallels Between Modern Physics and Eastern Mysticism” he explores the parallels between the odd-couple pairs of theoretical physics and Eastern mysticism. Although marketing ecosystems have not yet reached the intricacy of quantum physics (it’s getting close), marketers may benefit from the exploration of the resemblances between Eastern mysticism and modern MarTech. It’s this outside the box thinking that can lead to a more accurate understanding of how segmentation really works today in progressive marketing ecosystems.
Let there be segmentation!
“…an intelligent and personal God who stands above the world and directs it. Thus began the trend of thought which led, ultimately, to the separation of spirit and matter to a dualism which became characteristic of Western philosophy”¹
Traditional marketing segmentation techniques and strategies are a lot like Western philosophy; everything is neat and tidy; things fall into neat little buckets and you can name them. Often, they can even have fun names like “Sports Cars & Caviar” (for the more eccentric and affluent consumer) or “Strollers & Minivans”. With lovable names like these, marketers practically want to hug segments and can’t wait to bring them up at the next strategy meeting with exclamations such as “You won’t believe the response rates on our Skateboards & Hoodies segment — totally ill lift!”.
When marketers talk about “Gold” customers (or other elite monikers), they have a crisp image in their mind’s eye: currently engaged, transact often and make high value purchases. Unsurprisingly, these customers are the best customers to respond to offers — often, any offer.
Consider the three membership types offered by the Canadian Automobile Association, or CAA (AAA for Yankees or RACQ for Ozzies). Which of the three card holder types would you expect to be the most engaged segment for this roadside assistance company?
Premier CAA customers have already made the decision to select a membership type that is more than 2X the cost of a Basic membership. This is indicative of an existing established relationship and level of engagement with the brand where they are likely to respond to a variety of cross-sell offers such as additional insurance or participation in a rewards program.
It is no coincidence that segment definition closely resembles classic analytics and segmentation tools and models. “Marketers love RFM. I can’t get marketers away from RFM” says Mr. Uher in reference to the classic customer value model pioneered in the early days of database marketing which refers to Recency, Frequency and Monetary Value. “There is a comfort in the RFM model” he adds. This comfort is further reinforcing when marketers confirm that their “Gold” customer segment has brilliantly high RFM scores.
Being linear
“Classical physics was based on the notion of both of an absolute, three-dimensional space, independent of the material objects it contains, and obeying the laws of Euclidean geometry, and of time as a separate dimension which again is absolute and flows at an even rate, independent of the material world.”²
Traditional segmentation approaches are typically wrapped in within a linear workflow that is familiar to anyone who has looked under the hood of most direct marketing operations. The workflow is often some variation of the process illustrated below.
This approach seems perfectly natural within the Western Judaeo-Christian view of linear events: genesis, crucifixion, resurrection, the second coming and judgement day. It is easy to see why a sequential process like this would be instinctive for marketers. Traditional segments fit this model well, as they invite the definition of prescriptive buckets of customers that will fall into one segment before the next, and so on. If your job is to tell a compelling story in the boardroom, a linear story helps you to be understood.
There are pressures both within and outside of the marketing department that further reinforces this linear paradigm. Namely, privacy, legal and compliance are able to enforce checks and balances at agreed intervals. Similarly, quality control teams may insist on a series of approvals and validations at each step.
What’s not to love about elegant and linear processes you say? While easy to understand, many of these processes need to be reconsidered when personalized scale is introduced. Comprehensive manual quality control and legal checks are predicated on the possibility of “knowing all possible outcomes”.
When we take a traditionalist segmentation approach, the answer to “Do you define your segments in advance?” is yes; but, this manual approach quickly breaks down when we begin to pursue one-to-one personalization and we will need to seek inspiration from Eastern Mysticism for a solution.
We’re different. Our business is simple
Even simple businesses that have very few products or services will also encounter complexity that outgrows traditional segmentation when pursuing one-to-one personalization.
When we consider the possible permutations of unique customers, products, incentives and messages, the numbers add up. Take for example the case of CAA again, which has a relatively low number of products compared to say, a retailer:
This is what creating meaningful conversations at scale looks like, which are very difficult to orchestrate and maintain if you rely on a traditional linear approach alone.
Go East
“In the Eastern View…everything in the universe is connected to everything else and no part of it is fundamental.”³
Orchestrating 16 billion different possible interactions with customers calls for a different approach where we resist the temptation to predefine each possible outcome. It also necessitates working with contemporary tools, such as real-time dynamic decisioning, and marketing AI.
In the case of real-time decisioning tools, the marketer can only set the table for an interaction, but it is up to the customer to show up on an inbound channel (e.g. website, mobile app, call centre etc.) and drive the session. While you may predict in advance what a customer should want, the moment of truth arrives when that customer actually arrives and either explicitly or inexplicitly tells you what they want.
It is in this moment of interaction that a marketer’s well thought out static segments and predicted outcomes are required to yield, transform and interact with newly emerged context of the engagement. Marketers may seek inspiration to cope with this reality by studying the Kegon school of Mahayna Buddism, which is described as having a central idea to “grasp the universe dynamically whose characteristic is always to move onward, to be forever in the mood of moving, which is life”⁴.
If a customer is predicted to be interested in purchasing travel insurance, yet they are clicking and viewing pages related to driving education for their teenage son, the two competing segments require arbitration. Was this customer in the static segment for “travel insurance interest”?
Yes, but they also entered the segment of showing engagement for driver education soon after initiating their web session. These are examples of only two segments with their competing prioritization ebbing and flowing throughout multiple interactions, but there are many more micro-segments that make up the possible universe of interactions with each customer — none of these are “the” definitive segment.
Keeping the faith
If you try to change it, you will ruin it. If you try to hold it, you will lose it.⁵
Moving away from traditional segmentation techniques means changing the way your marketing group operates day-to-day. For professor Agrawal at the University of Toronto, this can change business workflows like marketing operations in two ways: “First, they can render tasks obsolete and therefore remove them from work flows. Second, they add new tasks”⁶.
In practice, this means a reduced effort for marketers on the development and reiteration of verbose marketing briefs with detailed segmentation definitions. Instead, energy is spent on describing guidelines and constraints within particular customer interaction moments. For example, what possible series of actions should we consider for a recently disengaged customer that has decided to revisit the website after a long hiatus? If you don’t detect any immediate urgent need, this could be the right time to remind the customer about the various rewards benefits associated with their roadside assistance membership, for example.
Transparency & causality
When it comes to adding new tasks, the adoption of artificial intelligence can often put marketers far outside their comfort zone, far beyond the familiar and reassuring RFM model. In part, this is due to the introduction of new tasks, such as the monitoring, calibration and application of constraints on machine learning solutions. This unfamiliarity can lead to uncertainty, as it isn’t always clear how marketing AI deployments will interact and complement co-existing traditional segmentation initiatives.
Aside from unfamiliarity, it is easy to be discouraged by some of marketing AI’s current unsolved challenges, such as a lack of transparency and false causality.
For the first issue, Agrawal describes transparency as one where “…AI technologies appear to be created from a black box” and goes on to explain that “it isn’t feasible to look at the algorithm or formula underlying the prediction and identify what causes what”⁷. This forces the marketer to change the way they narrate their success stories. Rather than touting their segment identification and targeting prowess, they can paint a picture of how they enhanced and refined a model that scaled to deliver outstanding results, while adhering to non-negotiable “do’s” and “don’ts” rules (e.g. exclude corporate members from non-applicable promotions or communications).
“I was surprised” said Mr. Uher when I asked him about overcoming concerns about transparency when deploying a new solution for a major retailer. Mr. Uher described a scenario where he was “able to totally diffuse the situation” and “they put away their pitchforks” by adding a marketer accessible means of managing manual customer exclusions. In this case, although marketers were not able to fully trace and explain the inner workings of the algorithm, they remained in full control of a “kill list”.
Examples of the second issue of false causality are wonderfully illustrated by Broussard at New York University, where she explains that machine learning can be highly effective, but for the completely wrong reasons⁸. In her experiment for predicting adult male Titanic survivors, her algorithm proves to be highly accurate, but reverse engineering indicates a strong bias towards favoring odd numbered lifeboats. However, it wasn’t the parity that was the root cause; rather, it was the fact that odd numbered lifeboats were on the port side of the doomed vessel, where the evacuation was under the direction of a different officer, who directed a more favorable plan for adult men.
Keeping the faith means focusing the evaluation of AI solutions based on their outputs and their performance, rather than attempting to reverse engineer the algorithms to conform to a more human understanding. To keep the faith, marketers need to accept that algorithms will “get things right” for “the wrong reasons” and focus their energy where they can add value: monitoring and confirmation of the results as opposed to trying to make an algorithm think like a human.
Achieving nirvana
When looking through the lens of current real-time decisioning and marketing AI tools, the question of “how many segments should we have?” by now should seem somewhat anachronistic. Yet, marketers cling to traditional approaches due to their familiarity and the need to tell easily understood stories to stakeholders and decision makers.
Whether you take inspiration from Eastern Mysticism or not, the boardroom is ready for new conversations and stories that reflect the reality of today’s tools. While traditional segmentation will never fully be replaced, keen marketers have much to gain by breaking the mould of the linear approach by introducing new stories about how they have used marketing AI and real-time decisioning systems to move the needle. It’s this evolution in storytelling that is a prerequisite for enabling meaningful conversations at scale, or marketing nirvana.
¹ Fritjof Capra, The Tao of Physics (Boston: Shambhala Publications, Inc., 1975), p. 20 ² Capra, 162 ³ Capra, 290 ⁴ D.T. Suzuki, The Essence of Buddhism, (The Buddhist Society, 1957) p. 53 ⁵ Lao Tzu, Tao Te Ching, chapter 29 ⁶ Ajay Agrawal, Prediction Machines (Boston: HBR Press, 2018), p. 126 ⁷ Agrawal, 197 ⁸ Meredith Broussard, Artificial Unintelligence: How Computers Misunderstand the World (Boston: MIT Press, 2018), p.116
