The AI Forecast that's Always Adapting

Tech Co. wanted to improve their in-house efforts to do demand planning

Tech Co. had a top-notch team of internal data scientists, but they were still looking for a third party demand planning solution that would help them make better business decisions

In particular, Tech Co. was looking for product that would to help them do three things:

1. Improve the accuracy of their demand forecasts to enable better planning
2. Forecast demand for new SKUs for which they had no historical sales data
3. Incorporate external event data (e.g., holidays, seasonality) into their forecasts

Core Demand Forecasting solution

• Created highly accurate demand forecasts for Tech Co.’s products at a granular level, including region, product category, and SKU
• Enabled sales projections up to 12 months into the future, with transparent error rates available to increase trust from stakeholders
• Incorporated seasonality, holidays, and promotions into forecasts to uncover key drivers of demand

New Product Introduction module

• Implementing solution to enable demand forecasting for new products by modeling similarity between new products and existing products in Tech Co.’s portfolio
• Designing demand forecasts to be updatable upon launch of new products to allow for adjustments to forecasts pending receipt of real-world sales data

Accuracy of retailer's existing forecasts was sub-optimal

Improving forecasts was made difficult by complex business hierarchies, shifting customer behavior, and high number of SKUs

Retailer had been using spreadsheet-based forecasts, which were slow and inflexible

These highly manual forecasts limited the business’s ability to adapt to market conditions in real-time

Retailer had no historical data for new products and new stores introduced each season

With limited data to inform traditional forecasting methods, this retailer was highly uncertain on how to plan for critical SKUs

Core Demand Forecasting solution:

• Built best-in-class demand forecasting using both historical & live data to continuously discover patterns
• Incorporated external event data to refine forecast precision
• Retailer now able to produce accurate, auditable, and adjustable forecasts that update weekly

New Product Introduction module:

• Modeled similarities between existing SKUs & new products to predict demand for new releases without direct historical data

Manufacturing Co.'s customers frequently make large orders of products with insufficient lead time

When orders come in with insufficient lead time, Manufacturing Co. is unable to produce and deliver on-time via standard distribution channels

When orders come in that Manufacturing Co.can’t fulfill by the requested deadline, there a substantial costs to the business

Manufacutring Co. must choose between two undesirable options: either (a) a delay in delivery, jeopardizing their business relationship with the customer or (b) an expensive express delivery via air, significantly impacting margins

Manufacturing Co. therefore needed a forecast of customers’ orders to ensure they were always stocked with the right products in the right places

This would allow Manufacturing Co. to avoid costly last-minute air transfer

Developing this forecast posed a signficiant challenge

The underlying data was sparse and Manufacturing Co. had previously just relied on human intuition to predict when orders would land

Core Demand Forecasting solution:

• Demand forecasts at the individual customer level enabling granular modeling of ordering behaviors despite sparse data
• Proactive alerts to sales team based on forecasts to ensure Manufacturing Co. is prepared to serve customers’ future orders before orders are even placed