Compare supervised and unsupervised segmentation techniques in marketing. Give examples.

Compare supervised and unsupervised segmentation techniques in marketing. Give examples.

• A. Supervised uses clustering to discover natural groups; unsupervised uses labeled outcomes to define segments.
• B. Supervised uses labeled outcomes to define segments only for acquisition; unsupervised uses forecasting models.
• C. Supervised uses clustering to group customers based on features; unsupervised uses rules based on expert knowledge.
• D. Supervised uses labeled outcomes (like churn propensity) to define segments; unsupervised uses clustering methods (like k-means) to discover natural groups without labels.

Answer: (D)

The main idea is that segmentation can be guided either by known outcomes you want to predict or by patterns you uncover in data without predefined outcomes. When you have labeled outcomes to predict—such as churn propensity, likelihood to respond to a promotion, or expected lifetime value—you build a predictive model using customer features. The segmentation then comes from the model’s predictions or the rules you derive from the model, grouping customers by similar predicted outcomes. For example, you might classify customers into high, medium, and low churn risk and tailor retention efforts accordingly, or segment by predicted campaign response to allocate creative and offers where they’re most likely to work. Without labeled outcomes, you turn to clustering to find natural groupings based on features like behavior, demographics, or past purchases. Algorithms such as k-means explore the data to reveal cohesive segments that aren’t defined by a pre-set label, like “loyal big spenders,” “occasional shoppers,” or “recent visitors with moderate spend.” These segments emerge from the data structure itself and serve as a basis for targeted messaging without relying on a specific predicted outcome. That’s why the correct description matches: supervised uses labeled outcomes to define segments, while unsupervised uses clustering methods to discover natural groups without labels. The other descriptions mix up which technique uses labels or clustering, or imply forecasting or expert rules where clustering on data would be expected.

The main idea is that segmentation can be guided either by known outcomes you want to predict or by patterns you uncover in data without predefined outcomes. When you have labeled outcomes to predict—such as churn propensity, likelihood to respond to a promotion, or expected lifetime value—you build a predictive model using customer features. The segmentation then comes from the model’s predictions or the rules you derive from the model, grouping customers by similar predicted outcomes. For example, you might classify customers into high, medium, and low churn risk and tailor retention efforts accordingly, or segment by predicted campaign response to allocate creative and offers where they’re most likely to work.

Without labeled outcomes, you turn to clustering to find natural groupings based on features like behavior, demographics, or past purchases. Algorithms such as k-means explore the data to reveal cohesive segments that aren’t defined by a pre-set label, like “loyal big spenders,” “occasional shoppers,” or “recent visitors with moderate spend.” These segments emerge from the data structure itself and serve as a basis for targeted messaging without relying on a specific predicted outcome.

That’s why the correct description matches: supervised uses labeled outcomes to define segments, while unsupervised uses clustering methods to discover natural groups without labels. The other descriptions mix up which technique uses labels or clustering, or imply forecasting or expert rules where clustering on data would be expected.