Setting up of Customer Segmentation
by Tom Wamelink, on Oct 26, 2020 2:07:25 PM
Segmenting customers can help you know, understand and follow your customer by finding groups with similar needs and characteristics. Furthermore it facilitates democratising the customers through your organisation to become more relevant to them in every contact.
In this article I will give an example of customer segmentation using a clustering algorithm, some thoughts on the process of getting there and tools that can be used for it.
Dealing with characteristics
You may have an idea of the stereotype customer that visits your company, but maybe this reflects just one characteristic. There will be more distinctive characteristics to find advanced and relevant groups. Although profiles like age or promotional buyer help you understand the size and importance of groups, it gives only one angle. Trying to combine multiple profiles ends up with many combinations leaving you with too small groups to target.
Segmentation is about finding customer groups that are distinctive to each other given a set of relevant characteristics.Your knowledge and experience of the industry is of great value and you can try different combinations of customer features. Eventually there can be many relevant solutions and you will succeed if you find the best one(s) for your use case. Many algorithms are available and the most known is probably clustering analysis using K-means, which is also the one that I use as an example in my blog.
Where to start
You can start with making a list of characteristics that are important to your customers and your business. These can relate to different dimensions like customer demographics, buying behavior, product usage, channel usage, needs states etc. For every dimension important to your business you can pick characteristics representing those dimensions.
For example a company in leisure may want to choose the moment of booking (buying behavior), the moment of arriving and accommodation (product) and the size and composition of the household (customer).
Example in leisure of segments in an optimum with eight features and six customer segments
Selection of features to form distinctive segments
Select the best feature(s) for every dimension and prevent features from interacting with each other (holiday spend amount and family size). Also consider your impact on the features so you are able to activate and differentiate in your customer segments. You can start with a first run with only a few important features to understand their distinctive power in forming segments. Note that distinctive power also depends on the selection of other features in the set. In next runs replace features without impact for other ones that you expect to be important. For the company in Leisure six segments were found in an optimum. Among two preferring bookings for campsites in high season, with one on a higher budget and booking more in advance than the other.
Running the algorithm
For running the algorithm, Python for example can be used that has libraries for clustering algorithms like K-means. I like the K-prototype algorithm (Huang (1998) )which is a great addition to K-means and overcomes the inability to deal with categorical features like accommodation type in the leisure example. If you are using the Google Cloud console then Google Data Lab can be used to run the cluster algorithm. Updating scores can easily be scheduled and stored in BigQuery tables.
Finding well formed segments is a great step towards personalisation. Getting there might take some time and multiple iterations, but it will enhance your view on your customers. When new customer characteristics are coming in this will give you new opportunities to segment your customers. Therefore consider it to be a continuous process helping you to understand your customer more and more.
A start can be simple with just a few characteristics and still find relevant groups. Once the algorithm is in place it can be rerun when updates of customer segments are needed.
Crystalloids helps companies improve their customer experiences and build marketing technology. Founded in 2006 in the Netherlands, Crystalloids builds crystal-clear solutions that turn customer data into information and knowledge into wisdom. As a leading Google Cloud Partner, Crystalloids combines experience in software development, data science, and marketing making them one of a kind IT company. Using the Agile approach Crystalloids ensures that use cases show immediate value to their clients and make their job focus more on decision making and less on programming.