Pluto7 successfully built the ML models to accurately predict the delay and quantity of the products

Introduction

Ulta Beauty has grown to become the largest U.S. beauty retailer and the premier beauty destination for cosmetics, fragrance, skin care products, hair care products and salon services. Ulta Beauty has been a visionary since day one. Seeing possibilities is what Ulta Beauty did when Ulta Beauty first created All Things Beauty, All in One Place — a store experience that connected with how beauty lovers actually shopped. And it forever changed the game.

Why Pluto7?

Pluto7 is an Artificial Intelligence (“Ai”), Machine Learning (“ML”) and Data Analytics Services Company with domain expertise in retail, manufacturing, hi-tech, automotive and healthcare verticals. With Pluto7 comes truly unique capabilities for helping higher education institutions leverage machine learning, artificial intelligence, and analytics on Google Cloud. Pluto7 helps industries keep up with ever growing demands on infrastructure, manage growing data archives, and helps in solving their problems with smart analytics solutions built on Google Cloud.

Solutions

The objective is to explore the historical data assets available in various repositories and identify key contributors required to predict supplier performance. Due to the pandemic situation they had a lot of problems. The entire process was manual to understand the supplier performance. As studied by Pluto7, There were 2 use cases to solve :

Use case 1: 

To predict the delay in delivery that can happen for the orders placed by ULTA to their vendors (suppliers). It was categorized as:

i) On time delivery

ii) (1-4) days delay

iii) (5-16) days delay

Use case 2:

To predict the quantity that will be delivered by the Supplier when the order is placed by ULTA.

Pluto7 is developing AI and ML Driven Operational Optimization using the best of breed products of “Google Cloud Platform” specifically tailored to Ulta Beauty’s requirements. The requirements mentioned in this document are based on the discussions with the Ulta Beauty team. Ulta Beauty can leverage these insights to drive and better focus on meaningful improvement opportunities aimed at reducing operational losses, optimizing delivery and providing incentives to suppliers for better performance using relevant KPIs.

Results

The data collected was uploaded in the Cloud Storage bucket. Multiple approaches of cleaning and transforming the raw data were done using the Vertex AI notebooks. The transformed data underwent multiple iterations, bigquery ML was also used to quickly come up with a baseline model on the pre-processed data. We built an accurate model for predicting delay in delivery with the accuracy of 72% and built another model for predicting received quantity with the accuracy of 88%.