A common marketing analytics challenge is to understand consumer behavior and develop customer attributes or archetypes. As organizations get better at tackling this problem, they can activate marketing strategies to incorporate additional customer knowledge into their campaigns. Building customer profiles is now easier than ever with BigQuery ML, using a technique called clustering. In this post, you’ll learn how to create segmentation and how to use these audiences for marketing activation.

Why are clustering algorithms so important?

Clustering algorithms can group similar user behavior together to build segmentation used for marketing. As we are breathing the personalization era, clustering algorithms can help companies send specialized messages to current customers or prospects through ads based on website behavior.

How do clustering algorithms work?

In this tutorial I’ll provide a simple understanding of clustering algorithms, however, the majority of this content will cover process and implementation, rather than what’s going on under the hood. To generally get you started, clustering falls under the category of unsupervised machine learning. We are running an algorithm, specifically in this process we will use k-means, to find how data is logically grouped together without giving the algorithm a target variable to train with. For example, suppose we want to cluster your audience by two characteristics like age and estimated income. Clustering is the process to automatically do this for you. The only input we’re faced with is how many clusters exist within our data. In the example below, three clusters ‘feels’ right. This example might seem straight forward, but you can see how the problem becomes impossible to manually do with more features.

#clustering #marketing #machine-learning #bigquery #advertising