Three years ago, at Sydney’s ‘CMO+CIO Leadership Symposium’, the CEO of IBM, Ginni Rometty, announced the death of customer segmentation.
The shift is to go from the segment to the individual. [Advanced analytics] spells the death of the ‘average’ customer.
Ms. Rometty’s core message was that we have the technology to communicate with customers on an individual basis. The rapid shift to tailored, individualized communication would render customer segmentation obsolete.
What is “Customer Segmentation”?
The term ‘customer segmentation’ is used to describe a broad spectrum of techniques for dividing customers into sub-groups – often for the sales and marketing purposes.
Simple segmentation may classify customers using attributes like age, or sex. More sophisticated, and more efficient, techniques group customers using both customer attributes and behavioral traits. For example, customers may be grouped by the way they interact with an organization and its products and services.
Behavioral classification is effective because it can distinguish between customers in ways that are easily actionable. Groupings may include ‘customers with decreasing spend’ or ‘customers who have made their first purchase of a product’.
Our definition also implies statistical and machine-learning concepts:
Customer Segmentation is a process, involving data analysis of multiple customer attributes and behavioral traits, that separates the entire customer population into distinct groups whose similar characteristics exceed the differences within the groups.
Our definition of customer segmentation has a number of features:
- It is data driven, rather than subjective.
- It distinguishes between customers across multiple attributes, including their behavior
- It segments customer across the entire customer base. It does not exclude customers for any reason
- The grouping process is not an absolute classification; some similarities and differences between customers can exist.
Classification is Not Customer Segmentation
Simple classifications of customers are easy to build. Unfortunately, simple classifications aren’t very effective. Simple classifications often group together customers with very different behaviors. This is insufficient for marketing purposes, or for analytical needs.
In real-world situations, segmentation using behavioral traits is much more effective than using customer attributes alone.
University Student Performance
We recently worked with a University that wanted to identify undergraduate students at risk of failing courses in the second or third semester of the year. The University’s goal was to identify and pro-actively support at-risk students. It also wanted to understand which students might go on to undertake postgraduate study.
The University’s initial segmentation used attributes like the course chosen and year enrolled. However, analysis showed that within these pre-existing segments, the number of students passing and failing courses was random.
A revised segmentation introduced behavioral traits. Measures of engagement with University facilities, like the library and online learning technologies, in prior semesters had a high correlation with future pass / fail rates. The inclusion of additional attributes, including demographic information, helped to further refine the segments.
The use of behavioral traits allowed the University to accurately identify students at risk of failing courses, and students likely to go on to further study. The Unversity was able to pro-actively support and guide both groups and create successful student outcomes.
Energy Demand Forecasting
An energy retailer wanted to understand and forecast energy demands across its customer base.
Existing demand forecasting processes used a self-reported classification of either ‘residential’ or ‘small and medium business’. In some cases these classifications were incorrect –customers labeled as ‘residential’ with thirty times greater consumption than the average were actually businesses. In other cases the broad classification was correct but did not reflect wide variations of usage patterns and consumption within the group.
Our approach was to use hourly consumption data to build intra-day and seasonal demand profile segments.
Placing customers into segments with similar historical consumption profiles radically improved forecast accuracy. Segments contained less noise and more accurately reflected complex energy demand patterns.
Target Retail Marketing
An FMCG (fast-moving consumer goods) retailer was getting poor results from their targeted marketing campaigns. They wanted help to re-develop their segmentation to improve campaign performance.
The retailers existing segmentation categorized customers based on lifetime spend alone. This segmentation classified all new customers in a ‘low value’ segment, irrespective of their spending patterns. Conversely, long tenure customers might be labeled as ‘high value’ even though they had not made a purchase in several months.
Our revised approach to segmentation implement ‘RFM’ – recency, frequency, and monetary behavior – extended by analysis of product preferences and product mix. This behavioral analysis established segments of customers who made purchases of economy or higher value products, reacted to promotions, or might consider purchases of specific product categories.
By targeting marketing campaigns with greater relevance to specific segments, the retailer was able to improve sales and revenue as well as campaign performance.
Is Customer Segmentation Really Dead?
When Ginny Rometty declared the death of segmentation, she also provided very insightful and accurate predictions of where customer analytics capabilities will be heading in the next few years. Effective one-on-one communication has the potential to revolutionise marketing, product and pricing strategies.
The problem for many companies is an effective customer segmentation strategy represents an aspirational future state. Using sophisticated analytical techniques, like the incorporation of behavioral analysis, businesses can improve response rates, sales, revenue and campaign performance.
Analytics8 has a great deal of experience implementing Customer Segmentation projects across many industries. We can help you deliver segmentation that creates business value for your organisation.