During the pandemic, e-commerce has emerged as the critical go-to-market channel. Many businesses have opened additional distribution centers and contracted 3PL services to adapt. In addition to these capacity-boosting measures, assessing safety stock levels is crucial to consistently meet increasing consumer demand while keeping an eye on working capital.
In the long term, maintaining consistent service levels requires revisiting – and perhaps revising – your forecasting models and modeling methodology. Your inventory deployment strategies will most likely require adjustment to maintain the safety stock level needed.
Forecast errors and the need for safety stock
We can all agree that every forecast has some level of error. But, imagine what you could do with a perfect forecast. In that case, how much demand-driven safety stock would you need? The answer, of course, is zero. (Although you still might need supply-driven safety stock and cycle stock.) In this ideal world, regardless of the volatility of demand, there would be no need for demand-driven safety stock. You would always know the exact level of demand to expect at any given time and how much inventory you would need to keep on hand down to the unit. This, of course, directly contradicts standard safety stock formulae.
Ideal worlds aside, in the here and now, forecasts are essential. They trigger demand and supply planning activities – setting expectations, providing targets, and mobilizing your organization to achieve higher fulfillment and service levels. Like many companies, you’re probably employing some of the sophisticated forecasting software that is readily available. Unfortunately, errors in implementation and execution often result in stock outs and, even worse, financial penalties and lost sales opportunities.
Every business needs to determine its safety stock requirements to manage customer service and their on-time and in-full (OTIF) and working capital requirements. There are two primary varieties of safety stock calculations: one that only deals with demand variability (the standard deviation of forecast error) and another concerned with demand and supply variability. There are other types, but these are the two most often applied. Working knowledge of these two is necessary for anyone responsible for inventory management. Safety stock is a necessary evil, and some would argue that it plays a vital role in the supply chain. However, as someone who spends a great deal of time forecasting, planning demand, and analyzing inventory, I’m often asked about the benefits of pursuing more accurate forecasts. One of the reasons is to anticipate demand and coordinate activities better; however, the degree of forecast accuracy rarely affects recommended levels of safety stock in most organizations.
Forecasting approaches and calculating deviation
Unfortunately, forecasts will contain errors. These errors represent the deviation of actual demand from the forecasted value – in short, the measure of how well a forecast performs. The smaller the error, the closer the forecast is to the actual demand. In this case, performance is a relative concept, and there are many measures of forecast performance.
The forecast error measure linked most closely to the standard deviation of demand is the root mean squared error (RMSE). The similarities between the two formulae are striking. Both involve statistical methods and variability analysis:
Standard Deviation of Demand
√(Σ(Actual Demand – Average Demand)²/(n-1))
Root Mean Squared Error of the Forecast
√(Σ(Actual Demand – Forecasted Demand)²/(n-1))
However, one striking difference is that the standard deviation of demand uses the difference between actual demand and average demand. In contrast, the RMSE formula uses the difference between actual demand and forecasted demand. If the RMSE is less than the standard deviation, replacing the standard deviation with RMSE in the safety stock formula would yield a lower recommended level safety stock. If the RMSE is more than the standard deviation, then use the standard deviation to calculate safety stock levels and the average demand as the “forecast.”
Safety stock approach impacts the bottom line
The implications of this are more far-reaching than just safety stock; it also provides the primary justification for forecasting. A $250M company achieving six inventory turns per year maintains nearly $42M in inventory. An improvement to seven turns, just one additional turn per year, reduces the level of stock to $36M. That’s a sizable reduction of $6M in working capital and a significant improvement of 14% in working-capital days. With results like these, forecasting can and should be the cornerstone of any lean initiative designed to limit over-production and inventories because the bottom-line impact is tough to deny.
At Chainalytics, we’re passionate about helping our clients through analytics, process, and technology. Our data scientists thrive on analytics. As the seasons turn and new products roll out with expectations for improved performance, an inventory Health Check may be precisely what’s needed to achieve your objectives. The Chainalytics Health Check reviews your forecasting software, processes, and results and identifies solutions for improvement.
In today’s demanding e-commerce market, your safety stock could make the difference between frequently disappointing or consistently satisfying your customers. Reach out to us and see how Chainalytics can help make appropriate inventory levels something you can rely on. Using one-of-a-kind tools and approaches like digital assets and managed analytics services, we consistently deliver actionable insights and measurable outcomes to our clients.
Hugh Walters is a Sr. Manager in Chainalytics’ Integrated Demand and Supply Planning practice. He is an expert in applying optimization, Six Sigma, and Lean principles to supply chain planning improvement initiatives.
