CSCA study materials
CSCA study resources

Safety Stock When the Forecast Is Wrong in One Direction

CSCA · Put your knowledge to work in context.

All case studies

Safety Stock When the Forecast Is Wrong in One Direction

A distributor of industrial filters has 24 months of weekly demand. Actual average demand is 800 units/week, but the statistical forecast has averaged 944 units/week for the last six months. MAPE is 22%, and the bias is not corrected in MRP. The supplier lead time is five weeks, with a standard deviation of one week. Weekly demand standard deviation is 180 units. The target cycle service level is 95% (z = 1.65). Current safety stock is 1,200 units, and the supplier MOQ is 2,000 units with a two-week order cycle. A four-week promotion is expected to lift demand by 25%, and no promotional forecast has been entered. Inventory is 3,400 units, and there are 1,100 units on open order. What should the materials planner do?

1. Diagnose the forecast problem and state the demand signal the planner should use for the next MRP run.

Read the worked answer

The forecast is biased, not merely inaccurate. Forecast 944 divided by actual 800 = 1.18, so the forecast is running 18% high. MAPE at 22% includes absolute error and does not reveal the direction, so MRP is being fed a consistently inflated signal. That builds inventory even while service may still miss because the promotion is not represented.

Use a de-biased baseline of 800/week for normal demand. Enter the promotion as a separate four-week forecast of 1,000/week, with a start and end date, rather than blending it into a smoothed average. Track forecast bias monthly as a separate metric. If bias stays above 10% for two periods, correct the statistical model or apply a documented bias adjustment. Do not simply raise MAPE-based safety stock; that would treat a directional error as random noise.

2. Calculate safety stock at the 95% cycle service level for normal demand and during the promotion. Compare with current 1,200.

Read the worked answer

Use safety stock = z x sqrt(LT x sigma-d squared + d squared x sigma-LT squared).

Normal demand: z = 1.65; LT = 5; sigma-d = 180; d = 800; sigma-LT = 1. Inside the root: 5 x 32,400 = 162,000; 640,000 x 1 = 640,000; sum = 802,000; square root = 895.5. Safety stock = 1.65 x 895.5 = 1,478 units.

Promotion demand: d = 1,000. Inside the root: 162,000 + 1,000,000 = 1,162,000; square root = 1,077. Safety stock = 1.65 x 1,077 = 1,777 units.

Current 1,200 is 278 units below the normal requirement and 577 below the promotion requirement. The planner should raise safety stock before the promotion and not wait for a stockout to justify it.

3. Given inventory 3,400, open orders 1,100, MOQ 2,000, and a two-week order cycle, what order should be released and what else must change?

Read the worked answer

Inventory position = 3,400 + 1,100 = 4,500. Normal five-week demand is 4,000; add normal safety stock 1,478 = 5,478 required, a shortfall of 978. The promotion adds four weeks x 200 extra units = 800, so the required position rises to about 6,278, a shortfall of about 1,778. With MOQ 2,000, release at least one 2,000-unit order now and plan a second order in two weeks if the promotion forecast is confirmed.

The bigger fix is process. The MOQ and two-week cycle create lumpy ordering, so negotiate a 1,000-unit MOQ or a scheduled delivery. Put the promotion into the MRP forecast with a start and end date, raise safety stock to 1,478 normally and 1,777 during the event, and review the order after week one of the promotion using actual sell-through.