Relationship Between Forecast Accuracy and Safety Stock
The amount of safety stock required by a business is directly related to demand forecast accuracy. In theory, a business could completely eliminate safety stock if your demand planner is able to predict sales for each product with 100% accuracy. At the other end of the spectrum – If you have zero ability to predict future sales, then you will need to hold a near infinite amount of safety stock to ensure order fulfillment.
Virtually every business realizes forecast accuracy between 20% and 80%. Almost all have at least some ability to predict sales in the future, and very few know with 100% certainty exactly how much of each product they will sell in future periods.
In this article, I will use some simple examples to illustrate how changes in forecast accuracy impact the required amount of safety stock. I will show you why 70% accuracy is often cited as an industry benchmark for forecast accuracy.
A Simple Model Shows the Impact to Safety Stock
I created a simple model illustrate the correlation between forecast accuracy and safety stock.
First, let’s assume your supply chain requires 60 days lead time. This is a reasonable starting point for this assumption – Most supply chains have at least some ability to alter supply plans two months out.
We will also assume you are targeting 30 days of safety stock. This reflects highly efficient inventory management, and is a good goal for a number of industries. Finally, we will assume your supply chain is capable of producing 125% of typical customer demand.
What Happens at 70% Forecast Accuracy?
Let’s say you are able to predict future sales with 70% accuracy. This means that if you sell 100 units of a product during a month, you would have forecasted sales between 70 and 130 units. For this analysis, we’ll take the low end of this range; we will assume the demand forecast is consistently 70% of the actual customer demand.
As shown on the above figure, we start the year with 30 days of inventory, which is our target safety stock. We actually sell 100 units, but only produce 70 units per our forecast. Thus, our inventory declines by 30 units, leaving us with safety stock of 21 days at the end of January.
The same scenario occurs in the second month – We actually sell 100 units, but consistently forecast 70% of actual sales. Because our manufacturing lead time is two months, we are not yet able to rebuild inventory lost in the prior month. Thus, our inventory again declines by 30 units, leaving us with 12 days of inventory at the end of February
March and Thereafter
We are still forecasting 70% of actual sales, so our predicted demand is still short by 30 units. However, we are now able to begin rebuilding safety stock we lost in January, since we have now satisfied the 60 day manufacturing lead time since that period. Thus, we produce our maximum capacity of 125 units (125% of customer demand). We recover to 20 days of inventory by end of March. We remain near this same level in future months.
The Results at 70% Accuracy: A Good Outcome
With demand forecast accuracy at 70%, our safety stock bottomed out at 12 days on hand of inventory, then recovered to 20+ days thereafter. At no time did we run out of inventory, and we didn’t exceed our ideal safety stock level. If you could manage this for each product variant in your portfolio, you would realize high inventory turns and virtually zero expired stock.
What Happens at 50% Forecast Accuracy?
Now, Let’s see what happens when forecast accuracy drops to 50%, all else being constant.
As shown on the above figure, at 50% forecast accuracy our inventory levels drop to literally zero at the end of the second month. This makes sense, if you think about it – We are forecasting one-half the actual demand, so we would burn through one month of inventory over a two month period due to forecast misses. Since we started out the year with one month of safety stock, then our inventory predictably goes to zero.
In this scenario, you would likely not be able to fulfill some customer orders at some point during the month of February. You eventually recover back to 15 days of inventory by April, but you are still running inventory well below your desired safety stock and customer orders are at risk.
Why 70% Forecast Accuracy is a Good Goal
Many manufacturers have a lead time of 60 days, and most would consider 30 days of safety stock an ideal level of inventory. Given these two inputs, a minimum of 70% forecast accuracy is needed to ensure the business doesn’t run out of inventory. For this reason, experienced demand planners often aspire to reach 70% accuracy in their demand forecasts.
What if I Can’t Reach 70% Accuracy?
While 70% seems like an achievable goal, the reality is that it is very hard to hit this level of accuracy on a consistent basis for most industries. Your entire sales forecast is likely above 70%, but it is very hard to predict sales for each and every product variant at this level. From my experience, I rely on strong support from sales and marketing teams, coupled with my sophisticated forecast models, to hit this level of accuracy.
Option #1: Hold Additional Safety Stock
Most companies opt for the easiest solution here – Simply hold more safety stock to accommodate variances between forecast and actual sales. In the below figure, I keep forecast accuracy at 50%, but increased safety stock from 30 days to 45 days. As shown here, this results in a minimum inventory level of 15 days, assuming forecast is always 50% of actual demand for a consistent period.
It’s never ideal to increase inventory levels, but this at least will allow you to fulfill customer orders until you can improve your forecast.
Option #2: Reduce Manufacturing Lead Time
If you cannot improve forecast accuracy, you can increase manufacturing flexibility to respond quicker to sales variances. You may choose to employ late-stage differentiation or other lean manufacturing processes to allow your supply chain to adapt.
By reacting quickly to changes in sales demand, you will be able to increase production before you run out of inventory. In this figure, I illustrate what happens with 50% forecast accuracy, but I reduce manufacturing lead time of 30 days. In this scenario, you never run out of inventory, as you are able to replenish safety stock the following month.
Demand forecast accuracy is directly related to the level of required inventory. Assuming 60 day manufacturing lead time, forecast accuracy of 70% will allow you to hold safety stock of only 30 days. As forecast accuracy declines from that level, you will need to hold greater amounts of inventory to accommodate unexpected swings in sales levels. As shown above, at 50% accuracy, your minimum safety stock increases from 30 days to 45 days of inventory.
Each subsequent decline of 10% of forecast accuracy will require you to hold an additional 5 days of safety stock. Thus, at 40% accuracy, you would need to hold 50 days of safety stock. At 30%, you would need to hold 55 days, and so forth.