Forecasting in Replenishment Supply Chains

Do Replenishment Supply Chains need forecasting? Or can they function purely on the basis of frequent demand signals coming from the market? It's a hotly debated topic with strong views on either side. Let’s try to understand the underlying reason behind this dilemma.

Replenishment Supply Chains use strategic inventory buffers at various decoupling points, which work as shock absorbers to meet sudden surges in demand. These buffers make the supply chain much more agile and responsive to fluctuations in consumer demand at the most granular level.

The buffer sizes are finalized based on estimated maximum demand during the Replenishment Lead Time (RLT). These buffers, however, must be resized periodically in a dynamic manner based on acceleration or deceleration of consumer demand. Frequency of resizing is one of the factors that decides the agility of supply chain.

Current level of demand, measured as Average Daily Usage (ADU) is generally used to calculate estimated demand during RLT, which is then multiplied by a factor based on variations in demand pattern to provide for peak demand. However, the fact remains that ADU is still an approximation for future demand, which holds remarkably true in a vast majority of cases and works much better than the monthly sales forecast coming through either S&OP or IBP.

There are certain situations, however, where ADU works as a rather poor approximation of future demand. These situations do prevail at times in a manufacturing company’s operations. Let’s examine three of the most common such occurrences.

1. Seasonal demand

There are several categories which experience seasonal variations in demand. Companies using Replenishment Supply Chain tend to revert to their monthly forecasts in such cases, with the usual inherent inaccuracies. On top of it, there is an added complexity of the season setting in either earlier or later, which makes the forecast even more unreliable.

2. Sales promotions

Brand owners offer various sales promotions from time to time to boost their sales. Sales uplift expected for each such promotion is estimated and used in forecasting. These promotions aim at disrupting the demand pattern and shape it in a positive way. As a result, ADU becomes a poor estimator of short-term future demand at the time of either getting into a sales promotion or coming out of it.

3. Month-end sales skew

Many companies link their salesmen incentives and performance measures to achievement of monthly sales. This results in a skewed uplift in sales towards the month-ends and much lower sales at month-beginnings. Replenishment models with inadequate smoothening read it as a case of demand acceleration as the month progresses, thereby pushing up the ADU. Demand smoothening does help in flattening it and maintaining a stable ADU. However, stocks still come under pressure at the month-end and start building up again when a new month begins.

These are the major reasons why replenishment models need to be tweaked from time to time. Many companies, who take the easy way of switching to monthly forecasts in such specific cases, suffer the inevitable consequences of poor forecast accuracy.

Is there a better way to deal with such cases in a replenishment model? Can we switch to Forecasted Daily Usage (FDU) instead of ADU?

I will cover it in my next article.