Franklin Sports deploys RL challenger models to prevent stock-outs across major league partnerships
A 79-year-old, family-owned sports brand deployed reinforcement-learning challenger models in seven days — turning Amazon point-of-sale data into a competitive weapon against stock-outs across MLB, NFL, NHL, NBA, NCAA, and MLS partnerships.
Franklin Sports was founded in Brockton, Massachusetts, in 1946 and is now headquartered in Stoughton, MA. The company manufactures and sells more than 10,000 products across professional, recreational, and youth sports — from the Pro Classic batting glove invented with Hall of Famer Mike Schmidt to equipment for MLB, NFL, NHL, NBA, NCAA, and MLS partnerships. With a primary shipping center in Memphis, an Asia office, and retail partners in dozens of countries, Franklin operates a global supply chain that must be as nimble as it is responsive.
Get the full case study
Download the PDF for the complete story and data.
What Franklin Sports needed to solve
Fast-fashion-style demand spikes
Championship runs, viral moments, and weather windows create sudden demand surges — like when an NFL team reaches the Super Bowl and fans need gear in time for the game. Serving league partnership requirements is, in effect, a fast-fashion supply chain.
Persistent out-of-stocks at major retailers
Retailers — Amazon in particular — are pushing automated procurement, but Franklin still faced gaps in out-of-stocks that hurt growth and strained customer relationships.
Need for insight at market speed
Scott Kennedy built an analytical framework around point-of-sale data, but needed a technology partner to turn channel data into actionable insight using reinforcement learning — with minimal latency as markets shifted.
Blended planning team across sales and operations
Franklin's demand-planning team spans sales-facing and operations-facing skill sets, requiring tools that surface probabilistic insight quickly enough for retailer conversations and account trips.
The approach
Scott Kennedy, VP of Digital Strategy & Analytics, worked with TrueGradient to build reinforcement-learning Challenger Models for Franklin's Amazon partnership. Using Amazon's point-of-sale forecast data, the models predict purchase-order requirements — giving Franklin's account team data-driven insight to challenge Amazon's forecasts and build a more collaborative, stock-out-resistant partnership.
Within a month, Franklin expanded the challenger model approach to other retail relationships. The team builds a probabilistic band of what could happen in the future, then collaborates with each retail account to align on future orders through reinforcement-learning-based modelling.
Franklin had already been working with Amazon's POS probabilistic models for years, so adoption was fast when TrueGradient built probabilistic forecasts at the wholesale level. The forecasts become the direction sales teams drive — enabling scenario planning for events like championship runs while grounding decisions in live point-of-sale data.
TrueGradient capabilities in action
RL challenger models on retailer POS data
Reinforcement-learning models trained on Amazon point-of-sale data predict purchase-order requirements — built in one week and expanded across retailers within a month.
Probabilistic account-level forecasting
Wholesale-level probabilistic forecasts replace point estimates as the basis for sales-and-operations alignment, with scenario modelling for promotions, events, and demand spikes.
Sales-and-operations collaboration workspace
TrueGradient's RL planner workspace gives Franklin's demand-insights team a shared tool to align with sales account teams on future orders at each retail partner.
Co-creation with TrueGradient founders
As an early adopter, Franklin helped shape the Reinforcement Learning Agent and Customer Success Agent inside the platform — with direct founder access rather than being routed to junior analysts.
Retailers are starting to provide automated procurement plans, but we still have a gap with out-of-stocks. The use of reinforcement learning, and the development of a predictive purchase-order plan, helps us prevent stock-outs, drive growth, and improve customer relationships. Working with TrueGradient made this easier because of the founders' experience in point-of-sale modelling. They were always there for me — I never got kicked down the organisation to work with a junior analyst.
— Scott Kennedy, VP Digital Strategy & Analytics, Franklin Sports
Measurable outcomes for Franklin Sports
Amazon challenger model built and in production within one week of kick-off
Challenger models expanded to additional retail relationships within one month
Improved retailer relationships and reduced stock-outs through collaborative probabilistic forecasting
Probabilistic forecasts adopted as the basis for sales-and-operations alignment across accounts
Franklin among the first customers to test TrueGradient's RL Agent and Customer Success Agent
More Case Studies
View all
Angelcare transforms Integrated Business Planning with Self-Serve AI
HexClad achieves 10% higher forecast accuracy with TrueGradient
Kisah achieved 2x sales growth during peak season with TrueGradient
Vastramay transforms pre & in-season planning with TrueGradient
Kapiva drives supply chain excellence with Self-Serve AI
Eggoz transforms S&OP with TrueGradient
Get clarity before your next planning cycle
Turn complex demand signals into clear, confident decisions without adding more tools or manual work.
For Shopify Brands
Optimize inventory, prevent stockouts, and boost profits.
Explore →
TrueGradient is a no-code self-serve AI product for supply chain optimization founded in 2023.
