Case Study: a leading multinational CPG company improves ML model runtime by 90% with Sigmoid MLOps

a leading multinational CPG company cuts ML model run time from 8 days to 14 hours with Sigmoid

A leading multinational CPG company faced significant challenges with its pricing and promotion machine learning model, which had an 8-day training cycle, high infrastructure costs, and was failing for multiple SKUs. This prevented migration to new geographies and caused missed business deadlines. Sigmoid was engaged to implement an MLOps solution to address these issues.

Sigmoid developed a multi-phase MLOps solution that automated model training and deployment, created benchmarking suites, and improved failing model features with a rule-based approach. This resulted in a 90% improvement in model runtime, reducing it from 8 days to 14 hours. The solution also achieved an 87% reduction in cost per run and a 20% improvement in MAPE, enabling the standardized migration of the model across multiple countries.