Beyond Single-Number Forecasting: Embracing Probabilistic Methods in Supply Chain Planning

There are two primary methods to make a prediction.

Single-Number Forecasting

The first method involves predicting that one specific event will happen.

For example, you might predict that the cricket team “Mumbai Indians” will win the IPL. Given that Mumbai Indians have been one of the most successful teams in IPL history, you might confidently place a single bet on them to win. In supply chain planning, this type of prediction is known as a ‘single-number’ forecast.

Planners, often using simpler systems like spreadsheets or legacy planning tools, forecast one number for a particular item. This approach can work well when you are confident that an established pattern will repeat itself, such as with fast-moving, commodity items.

For instance, if you have three years of history showing the sale of 100 standard USB chargers every week, forecasting 100 units is a safe bet. However, most products don’t follow such predictable patterns.

Probabilistic Forecasting

Let’s stay with the IPL analogy for a moment. Serious gamblers review a range of possible outcomes and apply their knowledge before making a bet. They may place multiple bets to hedge against losses from a single bet.

This scenario is analogous to probabilistic forecasting in supply chain planning. Advanced algorithms analyze multiple demand variables to identify the probabilities of a range of possible outcomes, pinpointing the most likely one. This method is more reliable for making predictions in situations where demand patterns are variable, historical order data is limited (such as with new product introductions), or factors like seasonality are at play.

Even if aggregate weekly or monthly demand for an item remains relatively consistent, daily demand at specific locations often shows considerable volatility. In a distribution network, examining aggregate demand is insufficient. To meet service levels, you need a plan ensuring the right number of items reach the right locations.

A probabilistic forecast accounts for uncertainty and helps manage risk. It improves average demand predictions and assesses the entire range of possible outcomes, including demand volatility, which significantly impacts service levels. With probabilistic forecasting, you still get one number associated with the highest probability. However, this number is surrounded by a range of other possible outcomes, each with a different probability.

The Benefits of the Probabilistic Method

Returning to the IPL analogy, betting on Mumbai Indians is likely safe, but where there’s little risk, there’s also less reward. Similarly, in business, there’s usually less upside in selling predictable commodity products. Most thriving companies profit from carrying ‘long-tail’ products in their portfolio, which command considerable margins.

Probabilistic forecasting is more than just a statistical method; it allows you to consistently make better inventory decisions for harder-to-forecast items than your competitors. This approach frees up working capital and improves service levels simultaneously, providing a sustainable competitive advantage that can elevate your company from good to great.

It also restores trust in your forecasting. When supply plans or safety stocks are based on inaccurate assumptions about demand uncertainty, targets are unmet, and supply chains enter firefighting mode. Trust in the planning process erodes. When planners lose faith in forecasts, they often hold too much safety stock, leading to excessive costs, waste, and obsolescence.

It’s better to hedge your bets with probabilistic forecasting.

Image Source: Stefan de Kok's post titled "A Primer on Probabilistic Planning and Forecasting"