AI Forecasts Beat the Incumbent Forecast in 15 of 15 Customer Deployments, DemandForecast.ai Benchmark Finds

Key Findings:

• Error reduction: forecast error fell by 14% to 56% relative to the incumbent forecast, with a 32% average across the 15 deployments.

• Planner effort: manual planning effort dropped by 70% on average. At Mars, 75% of forecast volume qualified as touchless.

• Overstock and sales: better forecasts cut overstock by up to 50% and lifted sales 10% to 25%. Nucor recovered $4M to $5M in annual sales at a single site.

• Portfolio size: deployments ran from a few hundred SKUs to more than 100,000, with horizons from 8 weeks to 36 months.

• Demand drivers: at one consumer hardware company, more than 20 assumed sales drivers were narrowed to four, with 80% forecast accuracy across SKUs and channels.”

The report names customers where they approved it, including Kenvue, Dorman, Mars, Nucor, Nanit and CAA Club Group. Each result is measured in the company’s own error metric. The 32% figure is the average of those relative error reductions across the 15 deployments.

“Every planning team has a forecast it has quietly stopped trusting, and very few ever get to see it tested against an alternative on their own data,” said Zohar Bronfman, Co-Founder and CEO of Pecan AI. “We ran that test fifteen times, on real data and in each company’s own metric. For supply chain leaders, the useful question now is how much accuracy their current forecast is leaving on the table.”

The full report, including methodology, limitations and downloadable charts, is available at demandforecast.ai/resources/forecast-accuracy-report

About DemandForecast.ai

DemandForecast.ai delivers production demand forecasting powered by Pecan AI’s Predictive AI Agent. It builds, maintains and keeps improving forecasting models on a company’s own historical data alongside its planning organization, starting with a proof of concept on that data. 

Contact

Head of Marketing
Ornit Rotenberg Haim
Pecan AI Ltd
[email protected]