Color coding retail data

The Benefits of Organizing Retail Datapoints using the Visible Light Spectrum 

Unique Challenge

Unify Consulting’s Delivery Manager for Data Science and Machine Learning Engineering Practice Dirk Biesinger identified a problem.  

Looking at how stores stock the shelves in retail locations, he wanted to figure out the best way to navigate the relative distribution of sizes for retailers. 

Making sure each location has the right number of units of each size for a specific product in stock presents a challenge.  

He sought to identify the best way for a retailer to use sales data to predict how many items in each size are needed, particularly when the number of sizes differs by product (some items have three sizes, some have four, denim products have an inseam and waist size). How can a retailer extract a typical profile, apply it, and make it comparable? 

Bespoke Approach

Biesinger decided to take the data and create a relative distribution curve and applied it to the visible light spectrum. This created a color code. 

He was able to identify a color code and, suddenly, it didn’t matter how many sizes a particular product had, because each one now had its own color code. Once Biesinger figured this out, he was able to use machine learning methods to create clusters in the data and create a comparison between the clusters.  

He tested the idea using a sample dataset of 10 million records compiled from 103 retail locations. He utilized machine learning to create clusters. He found the centroid of the clusters, the average value and stores clustered together. He took the information from the color codes and reversed it into the applicable buckets he needed. He was then able to work backward and create the relative distribution.  

Biesinger took the issue of various sizes out of the equation to build a tool that is useful to anything that is relative distributed along a linear axis.  

This process can be applied from everything to phones, features of technology products, reviews or sentiment scores, quality grades, etc.  

Proven Impact

Biesinger’s initial idea was sparked five years ago, and he has been building out the process ever since. He identified a practical problem and created a solution. And it is now a useful tool Biesinger applies to his work at Unify Consulting. Interested in learning more about Biesinger’s work? Contact Unify Consulting.