INDUSTRY: Oil and gas
SOLUTION: Supply chain and inventory management
TECHNICAL USE CASE: Data ingest and ETL, machine learning
For a large multinational like Shell, identifying ways to streamline the entire value chain across the globe is a continuous effort. Tracking inventory and ensuring maintenance issues is a daunting task. Today, they rely on Databricks for big data processing and machine learning at scale to dramatically reduce costs by improving supply chain operations and inventory management.
Keeping oil rigs and machinery operable is hard
To maintain production, Shell stocks over 3,000 different spare parts across their global facilities. It’s crucial the right parts are available at the right time to avoid outages, but equally important is not overstocking which can be cost-prohibitive. Their current processes and technology stack for maintaining inventory faced challenges.
- Disjointed inventory distribution: Stocking practices are often driven by a combination of vendor recommendations, prior operational experience and “gut feeling”.
- Limited DSS (Decision Support System) data availability: There has been limited focus directed towards incorporating historical data and doing advanced analysis to come up with decisions.
- Lost business agility: This can lead to excessive or insufficient stock being held at Shell’s locations, like oil rigs which have significant business implications.
Predicting equipment failures, improving inventory management
Databricks provides Shell with a cloud-native unified data analytics platform that helps with improved inventory and supply chain management:
- Databricks runtime: The team dramatically improved the performance of the simulations.
- Interactive workspace: The data science team is able to collaborate on the data and models via the interactive workspace.
- Cluster management: Significant reduction in total cost of ownership by moving to the Databricks cloud solution and gains in operational efficiency.
- Automated workflows: Using analytic workflow automation, Shell is easily able to build reliable and fast data pipelines that allow them to predict when to purchase parts, how long to keep them, and where to place inventory items.
Using ML to save millions on inventory management
Databricks is unifying teams across engineering and data science, allowing Shell to deploy machine learning models to predict equipment failures and inventory issues — improving operational efficiencies and dramatically reducing costs.
- Predictive modeling: Scalable predictive model is developed and deployed across more than 3,000 types of materials at 50+ locations.
- Historical analyses: Each material model involves simulating 10,000 Markov Chain Monte Carlo iterations to capture historical distribution of issues.
- Massive performance gains: With a focus on improving performance the data science team reduced the inventory analysis and prediction time to 45 minutes from 48 hours on a 50 node Apache Spark™ cluster on Databricks — a 32x performance gain.
- Reduced expenditures: Cost savings equivalent to millions of dollars per year.