- Customer segmentation is the division of customers into groups based on specific customer data.
- The aim of segmentation is to increase sales through personalized content tailored to the respective segment.
- Basis for any form of segmentation is centrally consolidated customer data.
- There are various methods of customer segmentation, including needs/value segmentation, clustering, RFM segmentation, predictive models, and segmentation at the individual customer level (segment-of-one).
Customers are like snowflakes: No two customers are the same. Even customers purchasing similar products or services have very different characteristics, needs, and interests. It is no surprise that customers differ in their demographics and locations, but traits such as product category preferences, average basket sizes, responsiveness, channel preference, and many more are often overlooked in marketing. Segmentation enables marketers to account for individual customer needs and their unique traits.
Segmentation is defined as division and grouping of existing or potential customers into segments based on their traits. This allows marketers to address a group of customers’ specific needs in a targeted manner. The highest level of segmentation is so-called segment-of-one marketing, which takes into account a single customer’s individual behaviors.
Customer Segmentation’s Goals
The overriding goal of customer segmentation is of course to increase sales. A better understanding of one’s own customers enables better personalization and customer experience, which in turn has a positive effect on customer interaction, conversion rates, and ultimately customer value and revenue.
Customer segmentation must distinguish between strategic segmentation and target group selection at the campaign level. Strategic segmentation has economic business objectives as its vanishing point and is used to identify initial or strategic target groups, such as a segment containing the most valuable customers. Although strategic target groups can also be the target groups of a campaign as a whole, they are typically segmented further for a campaign’s selection. For example, a segment containing the most valuable customers can be further broken down according to certain campaign-relevant criteria – be it gender, age, available channels, or the customers’ purchase histories. Objectives at the campaign level are more detailed and campaign success is measured by other key performance indicators.
For example, you may have identified a segment with loyal customers, but the value of that segment could be increased. The strategic goal for this segment would be to increase sales through better budget allocation and the key performance indicator would be the return on investment. At the campaign level, cross-selling campaigns can then be directed to sub-segments representing different interest groups. The aim of this campaign would be to lead customers to buy in a different product category and the KPI would be the campaign’s conversion rate.
Examples of strategic goals:
- Increasing turnover
- Increasing brand awareness
- Improving market positioning / differentiating from competitors
- Identifying new marketing opportunities
- Developing customized product offer strategies
- More effective allocation of resources
- Reducing churn rate
Examples of campaign objectives:
- Increasing repeat purchase rates
- Leading shopping cart dropouts to purchase
- Improving engagement/interaction
- Promoting a residual stock sale
- Communicating information about a new product feature
- Improving support for a specific issue
- Lead/address generation for an event
- Improving brand image after an incident
Data as the Basis for Customer Segmentation
Customer data forms the basis for every form of segmentation. Ideally, data is not stored in silos but is centrally consolidated, up-to-date, and can be immediately activated. If data, and thus the segments created from it, are not up-to-date, it isn’t possible to react to customer needs or behavior in time.
If data isn’t up-to-date, it isn’t possible to respond to customer needs or behavior in real-time.
Customer data is traditionally stored in a data warehouse (DWH). DWHs are powerful solutions for company-wide tasks in the field of Business Intelligence (BI). Their main task is data consolidation and standardization, which includes data analysis, data mining, reporting, and more. While it makes sense to consolidate customer data in a DWH, the functions of a DWH are not always sufficient for meeting segmentation requirements as segments are not made available quickly enough. The reasons for this are twofold: Firstly, DWH usage requires programming skills. A company’s IT team must be assigned to write queries and search the database. In addition, DWHs perform time-consuming ETL (Extract-Transform-Load) processes, which leads to significant latency between behavioral data tracking and the activation of this data by marketers.
Despite their inability to segment in real-time and the programming knowledge necessary to use them, it still makes sense to store customer data in a DWH and they are important parts of many companies’ infrastructures. However, for marketing and segmentation purposes a DWH should be complemented by a customer data platform (CDP). CDPs are generally specialized for usage in CRM and marketing. They store data in real-time and make it accessible to marketers through a visual user interface.
To learn more about the differences between DWHs and CDPs, read our blog post on the differences between CDPs and other solutions.
Find out what separates CDPs from DWHs and other solutions.
Customer Segmentation Methods
There are various methods of customer segmentation that need to be weighed against each other in terms of applicability and real benefits. They can also be combined.
Segments are created based on data in numerical form. Many attributes are numerical per se, be it data on purchase history, opening and click rates, or behavioral data tracked on a website. Qualitative data such as personalities, values, and attitudes first needs to be translated into quantitative metrics.
A customer base can be segmented one-dimensionally or multi-dimensionally. The former is segmentation according to a single customer attribute such as gender. However, a segmentation based on only one attribute often does not reflect reality accurately enough. Nevertheless, one-dimensional segmentation can be useful for creating segments quickly and easily. Multidimensional segmentation takes multiple customer attributes into account. Customer attributes can also be used as aggregated scores for segmentation, for example in the RFM model or CLV.
Needs/Value segmentation involves addressing specific customer needs in order to increase a value variable. Value variables can be, for example, the customer lifetime value, the average shopping cart value, or profitability. Segmentation begins with the attribute that has the greatest influence on the value variable and is then broken down more granularly.
Segment prioritization in needs/value segmentation can be based on experience and intuition, but loyalty levels of the customer groups or a scoring according to the RFM model can also be used to determine the customer groups’ values and compare them with each other.
The RFM model considers three core dimensions of a customer’s buying behavior: Time since last purchase (Recency), purchase frequency (Frequency) and sales (Monetary Value).
Until recently, the RFM model has been the gold standard for segmentation. It is increasingly being replaced by predictive models due to limitations, for instance, regarding reactivation campaigns and new customers: RFM scoring only dates back one or two years, meaning that it cannot help in reactivation campaigns for customers that have not purchased for longer than that time. New customers, meanwhile, have not purchased enough to have a clear RFM score. Nevertheless, it remains a valid method for identifying a company’s most valuable customers.
During clustering, unlike during needs/value segmentation, attributes are not grouped with regard to their effect on a specific target variable. Instead, it is performed without a fixed objective as it attempts to recognize patterns in the existing data sets. There are various algorithms that can be used to perform clustering.
Once segments have been formed via the clustering method they can be further analyzed and characterized to make them usable for campaigns. Hypotheses can help to better understand the clusters found and define their relevance for further campaign development.
Predictive models are mathematical calculations that predict customer behavior. As a rule, they use much more data than the RFM model – for example sociodemographic data, shopping baskets, the channel through which the customer was acquired, and other data can all be taken into account. This allows marketing strategies to be derived and budgets to be used more effectively.
CLV as a Scoring Model
The CLV is calculated based on the revenue and costs that a customer is expected to generate over the course of their entire customer life cycle. In other words, it is the estimated profit that can be generated with a customer throughout the customer life cycle. The CLV can be used to identify and target a company’s most valuable customers as well as those who are at risk of churn. In other cases, customer value can be increased through targeted campaigns.
CLV calculation can take various dimensions into account and the best formula depends strongly on the business model and purpose. Scoring can also weigh different dimensions differently according to specific requirements. Division into corresponding segments can be relative or absolute – a high-value segment can, therefore, contain either the most valuable X percent of customers or customers for whom the CLV exceeds a certain value.
Marketing budgets can be specifically planned for valuable customers and used, for example, to make them even more profitable through targeted cross- or up-selling. The CLV can also be used to assign costs and revenues to individual CLM phases and to plan budgets and campaigns accordingly. It is also possible to estimate which customers incentive measures such as vouchers are profitable for and how large these incentives should be. In addition, the CLV also brings advantages to customer acquisition. For example, it is possible to analyze which channels acquire profitable customers and then focus more strongly on these channels. It should be noted that the CLV is not a static value and can change over time.
The churn score describes how vulnerable a customer is to churn. This can be seen early on through a customer’s shopping baskets and purchase frequency. Ideally, factors that influence the churn score should be observed as soon as the customer relationship begins, so that engagement, product adoption, or loyalty can be worked on early on.
So-called “segment-of-one marketing” is based on automated segmentation and personalization in real-time. As the name suggests, segments-of-one are formed on individual customers’ characteristics or their behaviors. There are in effect two approaches to segmentation at the individual customer level, which can be combined:
- Segments-of-one by behavioral trigger: In these cases, customer behavior triggers a campaign in real-time or after a defined interval. With this rule-based real-time segmentation, channel selection and timing can be adjusted to individual customer preferences.
- Segments-of-one according to interests: Content is dynamically personalized in this approach, whereby customer interests are identified based on both their purchase history and their viewed pages/products. Examples include certain recommendation methods or onsite personalizations.
An example that combines both forms would be a shopping cart abandonment campaign: A customer adds products to the shopping cart but does not complete the purchase. After a defined interval, the customer can be automatically reminded of the shopping basket via the ideal channel, and the message can refer to the shopping cart’s contents. If the customer responds to the reminder, they can even be directed to a personalized landing page. You can test a shopping basket dropout campaign yourself here.
Segment-of-One marketing is the vanishing point for truly customer-centric marketing and customer relationship management. It enables relevant messages to be sent via the right channel at the most appropriate time, thus responding to individual customer needs. The prerequisite for segment-of-one marketing is a comprehensive view of individual customers and a real-time data infrastructure, such as that provided by CDPs.
Take a deeper dive into CDPs with our Beginner’s Guide.
How Many Segments Should You Create?
As a general rule, companies should only create as many segments as they can cover operationally. This is because practical limits are placed on segmentation by the available resources and systems used.
The number of segments at campaign level can typically be very high. Opinions differ on the ideal number of strategic segments: In theory, there is no upper limit as long as the segments gained are sufficiently differentiated and, above all, useful. The optimal number of strategic segments can also be determined using statistical methods, such as via certain clustering models.
As long as content can be played out automatically and does not have to be created manually, very granular segmentation can be achieved without much effort. For example, automated and individualized product recommendations in a newsletter are possible without large resource expenditure. It is therefore advisable for companies to think about how many segments they can use in practice before segmentation.
Not all of the resulting customer segments will be useful in practice. To evaluate the quality of a segmentation and the resulting segments, the following dimensions should be considered:
- Target-aimed: Does the segmentation contribute to general business objectives?
- Coverage of the customer base: Ideally, is each customer assigned to a segment?
- Identifiability: Is the segment sufficiently differentiated – does it provide recognizable differentiating traits?
- Size: Are the segments large enough to justify the effort and cost of personalization?
- Usability: Are the segments operationally applicable and do they provide a basis for effective campaigns?
- Accessibility: Can the segment be addressed efficiently enough?
- Stability/Dynamics: Is the segment stable over time or is it changing rapidly? Is the change taken into account if necessary?
- Responsiveness: Does each customer in the segment react in a similar way?
If customer segments are positively evaluated according to these factors and are used operationally, success must be continuously measured and evaluated in order to optimize strategic segmentation and campaign selection.
It is helpful to create personas for strategically important segments. Personas give a human face to data-based, impersonal customer segments. This enables marketers to better understand their customers and to address them with relevant and individual marketing activities.
Ideally, personas are based on the most significant differentiating factors between the respective target group and other segments. As personas are generic representations of target groups they only need to be created at a strategic level.
Customer segmentation is the basis for responding to individual customers’ needs, addressing them at the right time on the right channel with the right message. This requires customer data, which ideally should be centrally consolidated, up-to-date, and immediately activatable, so that customer needs and behavior can be responded to at the ideal times.
Each company should aim to perform segment-of-one marketing in combination with other segmentation methods. With increasing customer needs and expectations, companies must keep up with the customer experiences offered by the large platforms if they are to maintain and strengthen their direct customer access.