Setting safety stocks traditionally relies upon the random sums approach.
The data requirements for successful application of this approach, however, can be difficult to fullfill in certain settings and necessitate that accurate records be kept for determining the input parameters.
Furthermore, the random sums approach assumes independence between demand and lead time, a condition that could be
violated during demand spikes.
This paper examines the impact on safety stocks when record keeping is compromised.
An alternative formulation for setting safety stocks is proposed using a multiplication approach for estimating the variance that eliminates the need for pre-specified relationships and accounts for:
(1) data quality issues;
(2) demand and lead time correlation; and
(3) computationalsimplicity.
Using simulation, general applicability of the multiplication methoOrganizations typically make inventory decisions based on their
historical records. Data is collected over time and then analyzed to
determine inventory parameters such as safety stocks. The possibi-
lity exists, however, that the collected data has significant inaccura-
cies. In a report on the retail industry, Gruen and Corsten (2007)
note that poor data quality can be perpetuated by the disconnects
created in demand and sales data from not recording out-of-stocks
in the demand history file. Similarly, in an article on accounting
practices, Cagan (2015) maintains that small businesses often record
transactions only when money changes hands. Cagan also points
out that such a practice may even be encouraged as it provides a bit
of leeway at year-end to minimize taxable income. Issues of poor
quality daily demand data have also been observed by one of the
authors in the standard operating practices of a local distribution
center, stemming from resource constraints and resulting in peri-
odic demand reconciliation prior to placing replenishment orders.
Practices along these lines call into question the traditional statis-
tical approach used for managing safety stocks. Leveraging the fact
that the above recordation practices might lead to better data
quality when viewed at a higher level of aggregation than on a
daily basis, a computationally straightforward alternative for calcu-
lating safety stocks is presented herein.d is examined and the performance of alternative formulations is assessed. The multiplication approach, with its ease of application, is found to provide comparable results.