Safety Stock Calculation

For a simple explanation of safety stock, it can be said that the calculated safety stock is the difference between the sales forecast and the upper limit. This is in fact not true since there are more things that affect the calculations. Mostly it depends on how predictable the product is in sale, how well fitted the model is, the lead time, the cumulated confidence level for each lead time (normally distributed), the forecast and set service level.

In further details, safety stock calculations are used in addition to forecasts and confidence limits. This is most often used in setting inventories which are replenished only at certain variable or fixed intervals.

It is not very complicated to calculate the expected demand, but it is a lot more complicated to find the probability that the sales will exceed the cumulative forecast by some certain amount. This depends upon the details of the statistical forecast model. The difficulty of the calculation lies in considering the serial correlations of sales from point to point over the reorder cycle.

The solution to this is to convert the model to a form called the Wold representation. This is the key to determining the statistical distribution of the cumulative forecast. This does though assume that the statistical distribution of future sales is in fact correctly captured which is in fact never absolutely true since the model does not actually capture the true model.

The model will also assume it is equally likely for the demand to be above or below the forecast. To fix that upper and lower confidence limits provide the model with information about the spread around the forecast. This is important to know to calculate the safety stock. For example, a model that predicts 757 units in demand and has a 95% upper confidence level set as 875 units is in fact saying that there are 95% chance of demand being lower than 875 units. Note that the upper limit of sales forecasts is calculated from the set service level and variations in sale.

Those calculations do not take to account lead times but to determine economic order size and reorder point for some desired service level it is crucial to take lead time into account. The expected demand during lead-time (DDLT) is the cumulative forecast. Safety stock is the excess stock needed, above and beyond the DDLT, to maintain the service level specified for the upper confidence limit. The safety stocks are output for each lead-time up to and including the forecast horizon. Thus you would add together the DDLT and Safety Stock values for lead-time. This quantity is known as the Reorder Point. If your stock falls below the Reorder Point then you do not have enough stock to satisfy the expected demand at the specified service level and need to reorder.


Common Safety Stock Questions

1. Why is the safety stock on this product so high?

Rephrase the question: How do different parameters affect the safety stock?

  • Did you mean to look at minimum stock?

Safety stock is maintained to meet volatile demand during lead time (the time between placing an order and placing the units in stock). In other words, safety stock in AGR is added onto order proposals to make up for uncertainty in the statistical forecast.

There are three main factors to consider when looking at safety stock:

  1. Sale history: Statistical forecasts in AGR are based solely on each SKU’s sales history. The more robust sales history it has, the better the forecast will be. AGR’s forecasting module will pick up seasonal trends or other patterns if it has enough data points. If the sale history is very volatile or short, AGR’s forecasting module will not be able to produce a reliable forecast.

  2. Confidence factor: Every item in AGR’s systems has got a confidence factor which plays a big role in determining the safety stock. By applying a high confidence factor to an item, it is less likely that it will run out of stock.

    a. Side explanation: What does a 99% confidence factor mean? 95%? 50%?

  3. Order period: Remember that the safety stock is calculated relative to the order period. The item chart can be in days, weeks and months but the safety stock number will always be displayed as a horizontal line regardless of the time scale.

    a. Side note: Order period = lead time + order frequency

For further explanation, see these examples:

Example: Long lead time creates a seemingly high safety stock

Consider this item that has a long and robust sales history, high confidence factor (95%) and order period of 8 days (Lead time days = 1 and Order Frequency = 7) Forecasted quantities are generally around 2,000 units per month and safety stock for the order period is 81 units.

example-1

How does it look if the lead time was 6 months (lead time days = 180) but nothing else changed?

Safety stock goes up to be 1.025 units for the order period.

example-1-1

Example: Volatile sales history, high confidence factor and long lead time

This scenario is recreated from actual questions from customers.

Q: Looking at this item chart, the safety stock looks to be wrong. Monthly sales have never exceeded 1100 units per month, yet the safety stock is 1500? That looks like it’s trying to maintain nearly 6 weeks DC stock which we don’t want to do. Can you explain?

example-2

  • This is classified as an A product, so confidence factor is high (90)
  • Lead time is 130 days and Order Frequency is 30 days, so order period is 160 days (5.3 months)

Recalling that safety stock is a function of three main factors: Uncertainty in the forecast uncertainty, the number of days to cover (order period) and the confidence factor of the product. This item has got a mark in all boxes for a high safety stock:

  • High confidence factor (90)
  • A very long order period (160 days)
  • There’s very little sales history with a quick growth in the beginning and then decreasing just as quickly. This leads to high uncertainty in the forecast, which is drives the safety stock higher.

Is the safety stock high, or is it justifiable? Remember that the safety stock on the graph is calculated per order period. Rough calculations below can show us what the safety stock per day is and thus what the safety stock per month is.

Safety stock for order period = c.a. 1500 units.

Order period = 160 days.

Safety stock per day = 1500 / 160 = c.a. 9.4 units per day

Safety stock per month = 9.4 * 30 = 280 units a month.

The sales in July were around 1.100 unit and August were around 800 – what’s the sales in November (the next month) going to be? It wouldn’t be farfetched if it would sell 1.100 units again. By setting the confidence factor at 90, you are saying that you want to be more prepared for that to happen. Reflecting on the question: “That looks like it’s trying to maintain nearly 1.5 months DC stock which we don’t want to do”, it is not correct. It means that that the forecast might be wrong by up to 1.5 months of stock when ordering for roughly 5 months.

In other words, AGR is forecasting roughly 800 units a month but adds 280 units on top to make up for the uncertainty in the forecast.

2. How can I know what the safety stock is on my order?

See the order line info within the information sidebar within Inventory: 2.5 Orders.

order-line-info

3. What can I do to maintain a low safety stock?

Safety stock is a direct result of uncertainty in the forecasts. To see how you can influence the forecasts, see How to Affect Forecasts.

4. Is safety stock added to order proposals even when my AGR replenishing plans are not based on the statistical forecast?

Rephrase the question: For item ABC123, we did not like the forecast generated by AGR so we created a baseline plan for it that overwrites the forecast. Will AGR take Safety Stock into consideration before creating the order proposal?

Yes, it will be added onto the planned baseline quantity.

5. Can reports show ‘Safety Stock for Order Period’ based on scheduled orders?

This has to be created by an AGR consultant. You can be in touch with them through the AGR Service Desk.

6. What does a 99% confidence factor mean? 95%? 50%?

The higher a product’s confidence is, the higher the safety stock will be as AGR treats those items as items that should only go out of stock with low probability. To recall, AGR will generate the best fit statistical forecast using the items’ sales history but due to the nature of forecasts, it comes with some uncertainty. A very brief explanation of confidence factor is that it is the upper bound of the uncertainty.

These two products both have expected sale of 200 per month but with different certainty in the forecasting model.

Low uncertainty (Sales vary between 190 units or 210 units per month), safety stock is 10.

99-confidence-factor

High uncertainty (Sales vary between 100 units and 300 units per month), safety stock is around 40:

high-uncertainty

probability-chart

Although somewhat simplified, the graph is a recreation of the two screenshots from AGR above, where the forecasted value of both instances is 200 units but the safety stock is different. The blue line, which has a small uncertainty (low standard deviation), is highly concentrated around 200 meaning that the possible outcome is likely to be near that value, but could fall into some other value between 125 and 275 with at least some likelihood. The red line has a comparatively high uncertainty (a large standard deviation) meaning that while its most likely value is 200 it might also fall into a larger gap than that.

The shaded area under each line is 90% of the total area of the line. The blue horizontal line is at the value 210 and the red horizontal line is 240. Revisiting the screenshots above, that’s exactly what the safety stock is for those products. Take notice that their confidence factor is 90, thus explaining the 90% area. This means for the blue line that with 90% certainty, the expected sales will be less than 210. More generally, this means that by setting the confidence factor to a value, say value P, it means that with certainty of P-percent the forecasted value will be within the upper limit.

If the confidence factor was changed to 99, the blue horizontal line would move far to the end of the curve to roughly 275 units so that with 99% certainty, the expected sales will be within 275. In that case, the forecast is 200 values and safety stock is 75.

Confidence factors cannot be set at 100 since theoretically it would mean that the safety stock would have to be infinite.

If the confidence factor is set to 50, the horizontal line falls exactly on the forecasted value 200 so there would be 0 safety stock. This is ideal for products that have no uncertainty in their demand, for example made to order items.

Confidence factors between 0-49.9 are invalid since AGR is not concerned about anything that sells below the forecasted value.