Analysis of product package design elements through machine learning

Abstract

A product’s package is the first product element to be exposed to consumers and plays an important role in delivering information. However, prior research has been limited to human judgment and lacks systematic analysis of package elements. This study aims to systematically and quantitatively analyze the package elements used in front package design and identify typologies used in managerial practices. We extracted the features of the package elements by leveraging computer vision techniques and using real-life product package examples. Based on prior literature, we categorized all text and image format package elements. The features of the package elements were size, location, and color. The clustering algorithm revealed that four clusters are the optimum number for the data, with distinctive features: “Kids Appeal,” “Healthy,” “Classics,” and “Niche.” Finally, guidelines with specific numbers are suggested, to assist marketing managers in selecting the right package design elements to achieve the desired positioning.