Supply Chain Frontiers issue #32. Read all articles in this issue.
There is no such thing as a free lunch, except, perhaps, when it comes to counting inventory. Research being carried out by David Opolon, MIT Center for Transportation & Logistics (MIT CTL) PhD candidate, shows that a consistent inventory counting discipline can deliver productivity gains at no cost.
The research was originally part of a broader project to study the supply chain of an American mid-sized hospital in the Boston, MA, area. A common practice in such facilities is to use automated cabinets to dispense medical supplies. Authorized nurses access the machines, and take the supplies they need. Each time the withdrawals are recorded so the hospital can track inventory levels and demand.
Opolon was particularly interested in the accuracy of these recordings; for example how often did nurses actually press the required buttons to register that items had been taken. “I built a simulation model using realistic assumptions about the hospital’s demand patterns, replenishment schedules, and the behavior of nurses,” he said. Also, “how often the hospital counted inventory.”
The latter observations yielded some interesting findings. The data indicated that the hospital counted inventory once every seven days, but a more detailed analysis revealed that there were substantial variations in how often a product was counted in the same automated station. The average was once a week, but the overall pattern was erratic; item levels could be recorded once every four days or 10 days, for instance.
There were a number of reasons for the irregularities. A tally might occur if an individual saw that supplies for certain items were running low, or a manager had deemed that a count was overdue, are two examples. “On the one hand the variability could come from event-triggered counting strategies, but a large part of it came from a lack of defined counting policies,” Opolon said.
The hospital was paying a high price for these counting variations. “There was a significant loss of performance and wasted effort,” Opolon said. For example, the same service performance could be achieved by counting half as often but adhering to equally-spaced dates.
“Consistency brings you free improvement assuming it does not cost you to be consistent, and by and large it does not, it’s just a matter of setting a schedule and sticking to it,” Opolon said.
Why should unwavering counting schedules bring such returns? Counts that happen earlier than the target count date are less likely to find inventory errors. Counts that happen later than the target date will leave the inventory uncorrected for longer, possibly leading to stock-outs or unnecessary reorders.
When mistakes occur and the result is supply shortages organizations often react by increasing the frequency of inventory counts, only to see mixed results. There are a number of possible reasons. Human error is one; the more often inventory is tallied the higher the probability of a counting error. Moreover, increased count frequency lowers employee morale because this is an unpopular task. Before increasing the frequency of counts, managers should monitor the consistency with which counts are currently executed.
Opolon is now extrapolating the results to the retail industry, where cycle counts are relatively frequent at the distribution center and stock room levels. An important part of the model is forecasting the impact of different counting policies on the shelf availability of items. The model has four main parameters:
- The mean time period between counts (how often you count).
- The coefficient of variation of the time between counts (how consistently you count).
- The target service level assuming no inaccuracies.
- The level of inventory inaccuracy in the system relative to the level of demand uncertainty.
The model is simplistic, Opolon acknowledged, but it does provide a good estimate of the impact of counting policies on the probability of stock outs.
There is one important caveat: Variability could be a function of sound practices, for instance when there is a policy to perform counts on items that are past their due shipping dates, and which therefore are very likely to be out of stock. These practices should still be followed, since they are outside the routine cycle counting programs that Opolon’s model addresses.
In essence, reducing counting variation is an extension of the six sigma management strategy, which strives to improve quality by eliminating variations in manufacturing and business processes. “But this is an example of where you can actually quantify very clearly what the cost of variability is, and where reducing it is not prohibitively expensive,” Opolon said.
For more information on the inventory counting model and related research contact David Opolon, opolon@mit.edu .