PECO Pallet Improves Forecast Accuracy and SI&OP with Demand Analysis
PECO Pallet is committed to delivering quality wood block pallets and excellent service to their customer base of leading manufacturers of grocery products and consumer goods. Their extensive North American service network provides complete transportation coverage of the U.S., Canada and Mexico. PECO Pallet is headquartered in Irvington, New York, with over 1,600 pallet manufacturing, recovery, sort and full-service depot locations across North America.
PECO Pallet was challenged to generate an accurate network-level forecast. Experiencing 300% explosive growth over the last 10 years resulted in a dramatic change in their mix of renters, some with high-volume demand and contending with intermittent pallet returns from retail locations.
Their primary business objective was to generate repeatable, trusted process for month-over-month top-down forecast for upcoming 15 months to support sales, inventory and operations planning. They needed to improve their ability to predict and improve dwell time between pallet issue and return. Each day that dwell increased meant significant capital investment to maintain the needed pallet pool and honor committed service levels. Adding to this challenge, demand for each renter is impacted by different macroeconomic or external drivers that, without a tool like Demand Guru, was almost impossible to systematically quantify to generate a reliable forecast.
PECO Pallet adopted Demand Guru, LLamasoft’s demand modeling solution, utilizing clustering functionality, machine learning algorithms and external causal data to perform customer demand analysis and generate a network level forecast.
Because the vast majority of PECO Pallet’s supply and demand is driven by a selection of top customers, they chose to predict pallet returns by doing detailed analysis on a few key customers and using Demand Guru’s clustering capability for the remainder of customer volume. In addition to a resulting overall forecast, they separated out these large customers for individual forecasts.
They ran multiple models for each large customer and cluster group to determine what would drive future demand, applying different machine learning algorithms to hone the best result. Each group benefited from a unique set of techniques and external causals. For example, previous demand values, days in month, seasonal index, U.S. consumer price and GDP were key causal drivers for one large customer.
“Within two months of working with Demand Guru, we had a model in place that incorporated key external causals for 15+% improved forecast accuracy. This allowed us to generate better cross functional monthly planning and focus our energy on more detailed analysis.”
– John Solomon, Director, Transformation and Data Analytics, PECO Pallet
The customer has now made their new forecasting process repeatable by connecting with their live data warehouse. They use Demand Guru’s model output as part of their SOIP process and it is a key data point to set their supply/return for the coming months.
Learn more about Demand Guru and get LLamasoft’s latest eBook on AI-powered demand modeling.
Supersonic Market Sensing
A Breakthrough Guide to AI-Powered Demand Modeling