The impact of safety stock on schedule instability, cost and service

Abstract Frequent changes to production schedules in a material requirements planning (MRP) environment can lead to confusion at the shop floor level and reduced productivity. One approach suggested in the literature to reduce schedule instability is to introduce safety stock at the master production schedule (MPS) level to act as a buffer against differences in actual and forecast requirements. It is anticipated that, since MPS changes have a cascading effect on all lower level requirements and schedules, such a policy would help manage system nevousness. This paper examines the viability and effectiveness of using safety stock to reduce schedule instability at the MPS level. An experiment was designed to investigate the impact on schedule instability, lot-size costs, and customer service of using safety stock under a wide range of operating conditions. Five important variables characterized the environments under which the impact of safety stock was examined in this study: (1) the variation in end-time demand, (2) the uncertainty in end-time demand, (3) the item cost structure, (4) the length of the planning horizon, and (5) the lot-sizing method used. An intellegent safety stock policy was used which does not permit scheduling of MPS completions solely for the purpose of meeting safety stock targets. The criterion variables used in this study were schedule instability, lot-size cost error, and customer service level. The primary measure of schedule instability was a weighted average of schedule changes within the planning horizon that characterized the critically of changes by capturing the notion that it may be easier to react to schedule changes in periods of increasing distance in the future. In addition, a variety of unweighted statistics measuring the frequency and quantity of order changes were compiled for analysis. The cost error measure was a penalty cost computed as the percent deviation from the optimal total setup and carrying costs that would have been incurred by the Wagner-Whitin algorithm if demand over the entire study was known with certainty. Customer service was measured by the fill rate. The experimental results were examined using the analysis of variance procedure. The results indicate that increases in safety stock at the MPS level lead to higher customer service, but not necessarily to more stability in the production schedule or to a higher lot-size cost error. In general, the introduction of a small amount of safety stock improved schedule stability (compared to the alternative of no safety stock) and also lowered the cost error measure. However, further increases in the safety stock level often led to increases in schedule instability and always led to a higher cost penalty relative to the ex post measure of optimal cost. The increased nervousness of schedules was observed with a variety of alternative statistics that were used to measure schedule instability over the planning horizon. The results suggest that safety stock should be used with caution if it is introduced for the purpose of stabilizing schedules. In addition, the study identifies alternative strategies that may be effective in dealing with the schedule instability problem. For example, reductions is schedule instability consistently occurred when the expected time between orders was reduced and the forecast errors were low. This suggests that efforts to reduce setup costs and improve forecasting accuracy may be useful alternatives to increasing safety stock in dealing with the schedule instability problem.