Interpretation and its ambiguity
On the face of it, segmentation is a simple enough concept – divide the marketplace and use the differences between groups to meet a specific business objective. However, an array of machine learning algorithms and developments in clustering techniques mean the notion of “how to segment” can be interpreted very differently depending on the analyst.
Getting closer to the objective is key
Historically, segmentation was largely based on demographic data using variables such as region, gender and socioeconomic grade. This approach suffers from the over-generalization of large sub-groups of the population and as a result segmentation has shifted toward methodologies based on attitudinal and behavioral data. The advantage is these behavioral traits are directly actionable and can be incorporated into a model that separates consumers on more usable data.
However, to understand how to drive sales based on product strategy we must ensure the entire segmentation process is tailored specifically to this core business objective – rather than defaulting to purely attitudinal or demographic segmentations. In this example, it is vital to make sure the segmentation is based on product needs in order to understand what consumers want and why they purchase products.
Rise of machine learning
There are many machine learning algorithms available but broadly there are two fundamentally different approaches. Hard clustering assigns each consumer to one segment only, while soft clustering means consumers can be allocated to multiple segments, with a probability assigned for membership of each segment.
The advantage of soft clustering is its flexibility; it’s able to capture different elements of a consumer’s personality. A consumer’s behavior can change depending on factors such as mood or time of day. This technique’s therefore well-suited for products that are bought on a regular basis as its flexibility captures fluidity present in real-world decision-making.
When considering infrequent high-value purchases, for example buying a car, an algorithm that provides hard cluster assignment can provide a more reflective model of a consumer’s behavior. The binary, infrequent nature of these purchases means buying behavior is more consistent, making hard clustering highly predictive.
All these techniques have their place in segmentation but when designing a project it’s imperative to consider how a subset of the available methodologies and techniques can combine to create the most effective solution. What seems like a simple concept in principle can quickly become a very complex and intricate task.
DRG is an award-winning U.K. research agency group consisting of five specialist divisions. We provide innovative research solutions to some of the biggest brands, media owners, agencies and public sector organizations both in the U.K. and the wider world.
Our specialist in-house teams are experienced in analyzing data beyond standard data tables and use advanced statistical techniques and machine learning methodologies to unearth hidden insight to help drive key business decisions.