Data Analysis in Newly Developed Milk Sensor Platforms: Good Practices, Common Pitfalls, and Hard-Earned Lessons from Field Application

Abstract

In the last decade, the demand for healthier and safer food has increased alongside greater consumer awareness of food consumption, particularly in developed countries. This trend has pushed the food industry to implement a wide range of food quality control measures and surveillance systems for detecting contaminants. While high-end laboratory techniques remain the gold standard detection techniques, there is a growing need for simpler, more robust diagnostic tools that can be applied in the early stages of the food production chain to promptly identify deviations that may compromise food safety or quality. A complementary approach using both techniques can result in an enhancement of the overall contaminant-detection effectiveness and a better balance between food safety decision-making and the preservation of production value. This need is particularly relevant in farming and in the dairy industry. Developing milk process analytics requires careful consideration of both the nature of the processed sample and the conditions under which it is collected. Moreover, newly introduced techniques require the development of sound methodologies for data collection, analysis, and statistical process control. For this reason, this paper presents a detailed analysis of our previous milk data-collection campaigns involving technological prototypes, aiming to identify and suggest ways to preventively minimize issues related to experimental data collection, interpretation, errors, and mishandling. This analysis resulted in a set of practical observations and recommendations reported in the paper.