Marketing Analytics: Unsupervised Learning Model & Supervised Learning Model

Why is marketing analytics important? Mainly because, customers are demanding more personal and integrated services as they interact with marketing and sales through many channels. Most important of all, predictive marketing have proven to be useful by delivering values. In Some instances with enormous value! (ex. Netflix, Amazon, Facebook...etc)

Anticipating customer needs is not a new concept. The ultimate goal of a firm should be to anticipate and respond to customers needs automatically. To provide real time or near real time automatic service at large scale. Today's advancement in the technology infrastructure and with the increasing maturity of the AI technology has made this possible. Today, I'm going share 2 marketing analytic methodologies.

Unsupervised Learning (ex. clustering model)

Unsupervised learning finds hidden patterns in data, without explicitly trying to estimate or predict an outcome. For example, finding similar customers within a large group of customers, such as those who like long-distance running versus skiing, without explicitly knowing what groups exist or who belongs to them. Unsupervised algorithms such as clustering are thus typically used to unveil the true underlying segmentation of your data.

Artun, Omer; Levin, Dominique. Predictive Marketing (p. 25). Wiley. Kindle 版本.

At my current company we use clustering to unveil pattern regarding people that often purchase spare parts from us. The pattern of people that buy certain printing press from us. The pattern of the customers with high lifetime time value. The pattern of the customers that became a repeat customer...etc.

Supervised Learning (ex. propensity model)

Propensity model is also called the likelihood model. Supervised learning is used to estimate an output given an input, by training it with sample inputs and target. An example is to estimate the lifetime value of a customer, the likelihood a customer will engage with your brand, or a specific product a customer might want to buy next.

Artun, Omer; Levin, Dominique. Predictive Marketing (p. 25). Wiley. Kindle 版本.

At my current company we are using the propensity model. To input data from web behavior, email campaign interaction, demographics and industry category to predict the likelihood of the customer to subscribe, to become a member, and to purchase from us.