Part 3: Introducing Cloud Native Non-Forecast Based Inventory Optimization

As an early player in the forecasting and replenishment space for wholesale distribution, Thrive was amazed to discover that 80% of SKU’s sell 10 or less times a year per location for distributors. Thrive started to take a different approach based on 2 principles. Distributors needed a way to manage Unpredictability and Change.

1. 80% of distributor SKU’s are unpredictable. Too low of a volume and too much intermittent lumpy demand
2. At least 10% of distributor SKU’s sell to only 1 customer. If the customer is lost or the project is completed, the ERP and supply chain planning systems do not react fast enough to adjust inventory appropriately.

Thrive’s studies showed that 90% of dead stock results from these low moving SKU’s. Buyers have limited time each day so they naturally focus on the faster moving A and B SKU’s that comprise 80% of their company’s revenues. Noone has time to analyze each slow moving SKU.

No one also has the time to analyze each client loss or project completion. Additionally, the ERP systems and supply chain planning systems also don’t detect and provide fast enough insight into large vendor lead time changes (eg Covid).

Thrive’s AI platform addresses these issues with amazing success.

Real-World Results: When You Stop Forecasting

General Air Services: Increased Fill Rates 10 points on slow moving parts

General Air, a gas and welding supplies distributor, faced a problem that would sound familiar to any wholesaler: 97.8% of its SKUs were selling less than 10 times per year across 232,591 total active SKUs in 8 locations. Their manual, forecast-based process made managing this nearly impossible. They set an internal fill-rate target of 80% but consistently failed to reach it.

After implementing ThriveAI's Thermostock—which doesn't forecast at all but instead uses AI to generate optimal inventory targets based on actual sales patterns and business policies—the results were striking:

• Fill rate jumped from sub-80% to 88% consistently (with aspirations to reach 90%)
• The percentage of SKUs selling 10 times or less per year improved from 97.1% to 50.6%—not because sales magically increased, but because the AI identified which slow-moving SKUs actually needed to be stocked and at what levels
• Reduced dead stock while simultaneously eliminating lost sales

As Tara Cunningham, General Air's Purchasing Manager, put it: "The biggest difference after implementing Thermostock is that I do, 100%, have more awareness and understanding of how our inventory is stocked."

Digital Products Distributor: $29 Million in Overstock Eliminated

A digital products distributor with $217 million in annual sales was drowning in inventory—specifically, $29 million in overstock representing 44% of their total inventory. Traditional forecasting had led them to stock items that weren't moving, while simultaneously missing items that customers actually wanted.

Within just four months of implementing ThriveAI's approach, the transformation was dramatic:

• Overstock reduced from $29 million to $18.5 million (and continuing to decline)
• Overstock percentage dropped from 44% to 28.71%
• Dead stock at specific locations decreased from $9+ million to $5 million
• Company-wide dead stock reduced from $5+ million to $3.9 million

This wasn't achieved by better forecasting. It was achieved by AI-generated inventory targets that identified optimal stock levels for intermittent demand items and by SKU rationalization that identified which items simply shouldn't be carried at all.

The 45-Year Journey to the Wrong Destination

Let's trace the timeline:

• 1960s-1970s: MRP revolutionizes manufacturing with forecasting-based inventory planning
• Late 1970s-1980s: Distribution-specific ERP systems emerge (SX Enterprise, Eclipse) but inherit manufacturing's forecasting paradigm
• 1980: E3 Corporation founded to enhance forecasting capabilities for distributors
• 1990s-2000s: The industry consolidates around forecasting as the standard (Intuit acquires Eclipse in 2002, Infor acquires NxTrend in 2004, JDA acquires E3 in 2001)
• 2010s-2020s: Companies rebrand and merge (Activant and Epicor combine, JDA becomes Blue Yonder) but the fundamental forecasting approach remains unchanged

For 45 years, the wholesale distribution industry has been perfecting a solution to the wrong problem. The ERP systems got better. The forecasting algorithms got more sophisticated. The user interfaces improved. But the fundamental mismatch between forecasting methodology and distribution reality was never addressed.

The Bottom Line

The data is clear:

• Traditional forecasting was designed for high-volume, low-SKU manufacturing environments in the 1960s-1970s
• Wholesale distribution operates in low-volume, high-SKU environments
• Distribution-specific ERPs emerged in the late 1970s-1980s but adopted manufacturing's forecasting paradigm
• 80% of distributor SKUs sell fewer than 10 times per year
• Traditional forecasting methods fail for intermittent and lumpy demand patterns
• AI and machine learning methods can deliver 30-50% better profitability for these challenging scenarios

The pioneers who built Eclipse, SX Enterprise, and E3 Corporation weren't wrong to focus on wholesale distribution. They simply adopted the best inventory management practices of their time. But "best practices" from 1975 aren't best practices in 2026.

The question isn't whether distributors should abandon traditional forecasting—it's why they've clung to it for so long when the evidence has been mounting for decades that it doesn't work for most of their inventory. The technology has finally caught up to the problem. The only question is: how much longer will your organization keep using a 1970s solution to solve a 2026 problem?

The future of wholesale distribution inventory management isn't about perfecting forecasting—it's about moving beyond it entirely. Welcome to the post-forecast era.