Forecast Error, Lead-Time Variability, and Refrigerated Fill Rate
FreshCrate Foods sells refrigerated meal kits through 3 distribution centers. SKU-week forecast accuracy is 62% (MAPE 38%), and promotional weeks are worse. The company sets safety stock for a 95% cycle service level using a formula that assumes lead time is fixed at 5 days. Actual inbound lead time from the co-manufacturer ranges 4-11 days, with a standard deviation of 2.1 days. Promotions are planned by marketing 6 weeks out, but the co-manufacturer receives firm orders only 10 days out. Fill rate is 88% overall and 74% on promoted items; expedites cost $310,000 last quarter. The VP of Operations wants to raise safety stock across the board by 30%. Finance says inventory is already $14M and cash is tight. The S&OP manager wants a weekly collaborative forecast with the co-manufacturer but has no agreement on data sharing. You are asked to recommend a path to improve fill rate without a blanket inventory increase.
1. Diagnose why the current safety stock policy is failing, using the numbers.
Read the worked answer
The formula is missing lead-time variability. With mean lead time of 5 days and standard deviation of 2.1 days, the lead-time term can dominate. Illustrate with daily demand of 200 units, demand standard deviation of 80, and z = 1.65: ignoring lead-time variability gives safety stock of 1.65 × sqrt(5 × 80²) ≈ 295 units. Including it gives 1.65 × sqrt(5 × 80² + 200² × 2.1²) ≈ 752 units. The policy understates the requirement by about 2.5 times for that item.
A blanket increase misses the target. A 30% increase adds about 88 units to the 295-unit example, still far below 752, while over-buffering stable items that do not need it. The real drivers are lead-time variability, 38% forecast error, and promotions that are invisible to the co-manufacturer until 10 days out. Fill rate of 88% against a 95% target confirms under-buffering, but the fix is targeted, not uniform. Segment by lead-time variability and forecastability, then set safety stock with both demand and lead-time terms.
2. What S&OP and forecasting changes would you recommend, and how would you sequence them?
Read the worked answer
Change the commitment horizon, not just the forecast. The co-manufacturer sees firm orders only 10 days out while promotions are planned six weeks out. Extend the firm order window to three to four weeks for promoted items, with a flex band for non-promoted items. Share POS data and promotion plans under a data-sharing agreement; without that, collaboration is just a weekly phone call.
Segment the forecast. Use a collaborative forecast for high-variability and promoted SKUs, and a statistical forecast for stable SKUs. Measure accuracy and bias by category and by promotion, not just overall MAPE. A 62% overall accuracy can hide good performance on stable items and terrible performance on promotions.
Sequence the S&OP change. Pilot one DC and one category for 8–12 weeks, prove fill-rate improvement, then scale. Add a weekly demand review, supply review, and executive reconciliation. Do not buy a new planning system first; the constraint is the commitment horizon and data sharing, not software.
3. How would you decide between expediting and holding extra inventory for a promoted item?
Read the worked answer
Make the decision economic. Compare expected margin loss from a stockout with the expedite cost plus spoilage risk. For example, if a promoted item sells 1,000 units at a $3 margin and the stockout probability without expedite is 30%, expected margin loss is $900. An expedite at $0.80 per unit costs $800 and carries no spoilage. Expedite is slightly cheaper in that case.
Refrigerated items change the math. Extra inventory on a short-shelf-life item carries spoilage risk; a long-shelf-life item may justify safety stock instead. Set a rule: expedite for high-margin promoted items when the cause is a one-time event; otherwise fix the forecast and order horizon. Track expedite spend by cause. If the same SKU is expedited repeatedly, the problem is planning, not transportation. The $310,000 quarterly expedite bill is a signal to segment causes, not to ban expediting.