Product Features

• Single-Node and Multi-Node Supply Chain Replenishment Planning: Supports both single-node and multi-node networks, optimizing inventory management with minimal complexity for single-node and using advanced analytics for multi-node setups.


• Advanced Reinforcement Learning Algorithms: Utilizes AI models like Deep Q-Networks for simpler supply chains and Proximal Policy Optimization for complex multi-node environments to enhance inventory planning and demand management.


• Integrated Forecasting: Combines machine learning and classical statistical methods for accurate, reliable demand forecasting, adapting to large datasets with deep learning models.

Inputs, Configuration, Outputs

Single-Node inputs:

• Retailer Inventory Parameters: SKU, Selling Price, Inventory Capacity, Min/Max Order Quantity, Current Inventory, Lead Time, Stockout Penalty
• Historical Sales: SKU, Weekly Sales Quantity
• Sales Forecast: SKU, Weekly Forecast Quantity
• Order In-Transit: SKU, Order Date, Order Quantity

Multi-Node inputs:

• Distributor Center Inventory Parameters: SKU, Inventory Capacity, Min/Max Order Quantity, Current Inventory, Lead Time
• Retailer Inventory Parameters: Retailer, SKU, Selling Price, Inventory Capacity, Min/Max Order Quantity, Current Inventory, Lead Time, Stockout Penalty
• Retailer Historical Sales: Retailer, SKU, Weekly Sales Quantity
• Retailer Sales Forecast: Retailer, SKU, Weekly Forecast Quantity
• Retailer Order In-Transit: Retailer, SKU, Order Date, Order Quantity

Configuration: The capability offers a versatile set of customizable options, allowing businesses to adapt inventory management and forecasting to their specific needs.

• Forecast Flexibility: Choose between user-generated or system-generated forecasts.
• Forecast Length: Select length of the forecast and planning period in weeks to match operational timelines.
• Adjustable Week Start and End Days: Allows users to set the start and end days of their week.
• Order In-Transit: Includes an option to account for order in-transit information in replenishment calculations.
• Simulation Options: Users can specify the number of simulations to run, ranging from 100 to 100,000.
• Maximum Runtime: Set the maximum runtime for simulations from 1 to 24 hours.

Outputs:

• Inventory Start / End Positions: Shows the inventory position at the beginning and end of each planning period.
• Demand and Sales Metrics: Includes total demand, units sold, unfulfilled demand, and demand fulfillment rate.
• Inventory Management: Details orders arriving, in-transit inventory arriving, projected future inventory, future inventory shortfall, and excess inventory.
• Actions and Adjustments: Records original and adjusted actions taken, along with reasons for adjustments.
• Performance Metrics: Provides overstock amounts, capacity utilization rate, and an overall reward metric indicating the efficiency of the inventory management.

Technical Architecture

Hybrid Cloud Architecture: Split Control Plane (hosted by MetaFlow) and Data Plane (hosted by Customer, managed by MetaFlow).

MetaFlow cannot directly access the resources in the Customer Cloud account and interacts with them only through AWS APIs. Customer data stays entirely within the Customer Cloud account/VPC, both at rest and in transit.

Example Workflow:

• Deployment: A new version of MetaFlow service is deployed via the Control Plane. The Control Plane manages the deployment by instructing the ECS cluster in the Data Plane to pull the new container image from AWS ECR.
• Integration: Users can integrate MetaFlow Service with their existing Supply Chain applications using ETL pipelines, securely accessing their AWS S3 storage for seamless data transfer and processing.
• Data Processing: The deployed service in the Data Plane processes the customer supply chain data, running AI models or analytics scripts as required.
• Monitoring and Scaling: If the load on the processing service increases, the Control Plane detects this and automatically scales the ECS service to handle the additional load.
• Result Delivery: Processed results are stored in the customer AWS S3, and the Data Plane updates the customer dashboard with the new insights.

Setup and Integration Process:

• Role-Based Access Control: MetaFlow assumes specific roles in the customer AWS account for API actions, without accessing data directly.
• Provisioning of Resources: MetaFlow uses Terraform to set up required resources, including ECS clusters and load balancers, in the customer cloud.
• Security and Compliance: Ensures data remains within the customer VPC, with Control Plane interaction limited to AWS APIs.
• Monitoring and Management: Monitors service performance, with automated recovery processes in place for any detected issues.

Implementation Services

To ensure the product is correctly set up and operating effectively, MetaFlow offers two tiers of service:

• Standard Service: Includes email support to assist with setup and any technical inquiries during implementation.
• Premier Service: Provides dedicated support with a dedicated team to assist with custom setup and ongoing support throughout the implementation process.

Solution Screenshots

×