How cluster analysis identifies market and customer segments

When you don’t research and segment your market and customer base, time and effort may be wasted on product development or campaigns that simply don’t resonate with your potential customers. 

SurveyMonkey offers a Consumer Segmentation solution that will help you build your buyer personas, measure product demand and guide campaign targeting. The solution includes cluster analysis, which is an integral part of identifying both consumer and market segments. Let’s take a more detailed look at cluster analysis and the role it plays in segmentation.

Cluster analysis is a data analysis technique that identifies meaningful, naturally occurring groups within a dataset and distinguishes them as clusters. It is used to discover hidden relationships in data based on specific characteristics.

Can you manually segment your customers and market? You certainly can, but the manual method is limiting and only truly effective with a small number of characteristics or attributes. It simply doesn’t scale well. Cluster analysis with advanced analytics and machine learning can quickly scale to a high number of attributes. It’s also completely data-driven, using an unsupervised model (the algorithm learns patterns without tagged data), which makes it more accurate and credible.

The hierarchical clustering technique is a commonly used, easy clustering technique. It is divided into two types: 

The basic algorithm does the following:

The basic algorithm does the following:

Both cluster analysis and market segmentation involve grouping customer segments based on similarities. While segmentation is based on human input, cluster analysis is driven by machine learning. Cluster analysis provides insights that allow businesses to drill down into the needs and wants of each market segment, allowing them to offer more personalised products and messaging. Using cluster analysis, you can identify new target market segments as well as ones to avoid.

Your clusters in market segmentation will usually have a heavier emphasis on geographic information, such as metropolitan areas, counties, countries, and regions, etc., and demographics, such as age, income, and gender, etc.

Examples of cluster analysis in market segmentation

A company has created what they consider to be the perfect cocktail dress. They have it priced at £750. They want to target the appropriate market for the dress, so they know they need to find people who can afford it. Traditional segmentation might be based on their belief that only people over 45 have the income to purchase their dresses. But they used clustering and found that younger women aged between 25 and 35 not only have the income but are also more likely to purchase cocktail dresses. This had a significant impact on their marketing strategy.

You own a business that provides lawn care products to your overall target market of homeowners. You’ve collected information via a survey that includes household size, income, primary shopper and distance from the nearest town/city. By entering these variables into a cluster analysis tool, your results reveal clusters based on the size of the family and their spending (e.g. small family + low spending, small family + high spending, large family + low spending, large family + high spending).

Cluster analysis in customer segmentation is used to create homogeneous groups of customers. In general, customer segmentation is used to identify behaviours and attitudes of the groups that you’ve segmented by market. Cluster analysis will reveal clusters based on these characteristics.

Examples of cluster analysis in customer segmentation

You’re ready to launch a new streaming entertainment service. To ensure that your marketing is focused on the right consumer segments, you conduct market research to find out basic demographics, how many minutes per day are spent viewing content via streaming, how many days per week include watching content via streaming and the number of unique shows viewed each week. Cluster analysis reveals clusters that identify high streaming usage, so you can plan your marketing and advertising based on the consumers who are most likely to be interested in your service.

Your email marketing campaign isn’t performing as well as you’d hoped it to. You collect information about your recipients, such as how many emails they have received, how many they have opened, time spent viewing the email and the number of clicks per email. Using this information, cluster analysis can identify and group recipients who interact with your emails in the same ways. You can then customise your content and email frequency based on the characteristics of each cluster.

Cluster analysis uses machine learning algorithms to group things – in our case, customers – into similar groups or clusters. There are two main algorithms used in clustering: K-means analysis and K-medoids analysis.

The K-means algorithm divides a single cluster into K different clusters. It does this by finding organically similar data points and assigning each one to a cluster with similar characteristics. K-means clustering works by constantly trying to find a centroid (a data point that represents the mean or centre of the cluster). The end clusters will each have a centroid and data points that are closer to the centroid compared to the other centroids.

There are pros and cons associated with using K-means cluster analysis:

Pros

Cons

K-means cluster analysis is widely used across several verticals, from determining urban traffic patterns for Uber drivers to segmenting customers based on interests, purchase history or buying behaviours.

The K-medoids algorithm is similar to the K-means algorithm, but instead of using centroids, it establishes medoids (the least dissimilar data points). The medoid is an actual point in the data set, whereas the centroid in K-means is an average (mean). K-medoids cluster analysis is more accurate because it is not sensitive to outliers like K-means can be.

There are pros and cons associated with using K-medoids cluster analysis:

Pros

Cons

The K-medoids algorithm is used in facial recognition software because it uses an actual data point and is more robust to outliers. If K-means were used, it would be less effective because there is no real data point to start with but rather a centroid with mixed features from several photos. 

In some business and marketing cases, K-medoids cluster analysis would be used because it is preferable that the centre point is an actual data point. 

Cluster analysis is used in market research for a variety of reasons. As we discussed earlier in this article, it’s particularly useful in developing market and customer segmentation. With cluster analysis, you can:

Although cluster analysis is used primarily for segmentation, the results are applied in various ways in marketing strategies. You can use the data to:

It’s time to look at a practical application of cluster analysis, based on a case study from Momentive, the creator of SurveyMonkey. The study, which is titled ‘Momentive study: credit card users weigh costs and perks’, used cluster analysis to group credit card users. This allowed researchers to create groups that were easy to understand based on shared behaviours and preferences. These coherent groups allowed an increased understanding of the different behaviours and preferences among the broader group of credit card users.

Using cluster analysis, we were able to create four unique segments that represent types of credit card users. The data inputs to reach these segments include basic demographics, such as income, gender, race and age, and usage and attitudes towards credit cards and credit card features.

This consumer segment, which is defined through cluster analysis, skews those who are older and risk-averse to using credit cards. Card ownership is high in this segment and they generally go for cards with no annual fee. They also focus on maximising cashback and other passive card perks. This segment tends to pay off credit card debt in full each billing cycle.

The Untapped Utilisers segment identified by our cluster analysis tends to be lower income. They are the least likely of the four segments to own a credit card and are relatively unfamiliar with the benefits. If any, they will have one or two cards that are likely to offer cashback. This segment shows little interest in maximising card rewards or other benefits.

Younger consumers – Gen Z or millennials – have less disposable income but frequently use credit cards. This segment is willing to apply for store or co-branded cards with an annual fee. Strapped Spenders most often pay the minimum payment for each billing cycle. Strapped Spenders are yet another segment identified by cluster analysis.

Our cluster analysis defined this segment that mostly comprises high-income millennials. They are interested in maximising card benefits, especially travel, holiday accommodation and airfare bonuses. Card Champions will pay high annual fees for premium credit cards in order to earn the best rewards.

The four key findings in the Momentive study are as follows:

Without cluster analysis, we would not have been able to identify our four unique segments of credit card users. These segments provided us with unique insights into each one, including what features are most important, how younger Americans are using credit cards and the reasons for applying for new credit cards.

The identification of these segments can serve essential roles for credit card companies in the future. Depending upon which segment is the target market for a particular credit card, marketing can be targeted to their particular pain points. For example, if a company wants to target the Untapped Utilisers, marketing and advertising can focus on cashback rather than other benefits or rewards. But if the target segment is Card Champions, messaging should focus on high-level benefits such as travel, holiday accommodation or airfare bonuses.

Cluster analysis is an effective way to identify market and customer segments for your business. The data can be used in a wide variety of industries to inform marketing strategies. In the Momentive case study, cluster analysis identified four consumer segments to help credit card companies identify their target segments and tailor their marketing efforts towards their desired segments.
If you need help with your consumer segmentation, our segmentation tool is just for you. We offer a variety of other market research solutions as well. Let’s get started!