Principle (Variability Buffering): Variability in a production system will be buffered by some combination of inventory, capacity, and time.
Motivation
All processes contain some variability. In most manufacturing environments variability degrades performance. For example, the best case performance of the closed infinite buffer station occurs when there is no variability. The practical worst case occurs when variability (in process times) is quite large. One method for reducing variability in certain situations is variability pooling. We can also reduce variability in process times directly by improving reliability, reducing setups, enhancing training, etc. But regardless of what we do, some variability will remain. And by this principle, that variability MUST be buffered. The following examples illustrate how inventory, capacity and time can act as buffers against variability.
Examples (Production)
Bottleneck fed by Unreliable Process: Consider a production
system in which a bottleneck process is fed by an upstream station that
is subject to failures. The variability here is the arrivals to the
bottleneck, which are disrupted by the failures. If we want to keep
the bottleneck running a high percentage of the time (i.e., not
use capacity as a buffer), then we must maintain a WIP buffer in front
of the bottleneck. This is a classical production example of an inventory
buffer.
Bottleneck fed by Variable Speed Process: Now suppose that the bottleneck is fed by an upstream station whose speed varies (e.g., depending on the product being run). Over the long-term the upstream station is faster than the bottleneck, but there are intervals when it can be slower. One way to ensure that the bottleneck keeps running is to implement a WIP buffer as in the above case. However, an alternative would be to increase the capacity of the upstream station. This would enable the station to keep up with the bottleneck even during slow intervals and would thereby reduce or eliminate the need for a WIP buffer. In cases like this, capacity and inventory buffers can be alternatives to one another. Which is best generally depends on costs (i.e., cost of capacity, cost of holding inventory, cost of increased flow time due to increased WIP) and will vary from system to system.
Overtime in JIT: Just-in-time systems try to stay away from using WIP as a buffer. They also try to minimize demand variability via production smoothing. But even in the best run systems, the production rate will be variable to some extent. To match production to demand, JIT systems typically make use of a periodic production quota (e.g., an automobile assembly plant has a quota of 400 cars per shift). If such a system runs three shifts per day, seven days per week, then any quota shortfalls will translate into delays in filling customer orders (i.e., a time buffer). Since delays are bad from a customer service standpoint, JIT systems typically build in a capacity buffer. One way to do this is by simply setting the production quota safely below the capacity of the system. Another is to schedule two shifts per day, separated by a four hour preventive maintenance (PM) period. If a quota is not met, then overtime is used during the PM period. Alternatively, we could schedule five days of production and use the weekends to make up quota shortfalls. Either of these represents a production example of a capacity buffer, since we are deliberately scheduling below capacity.
Due Date Quoting: Consider a wafer fab. Excess capacity is
extremely expensive, which makes a capacity buffer unattractive.
Inventory loses value rapidly, which makes a WIP or finished goods buffer
also unattractive. Hence, in such systems, due dates are typically
negotiated (as opposed to quoted uniformly in advance) to account for variability.
For instance, a customer who places an order when the plant has a heavy
backlog will get a long lead time quote. A customer who places an
order during a lag in demand will get a short lead time quote. This
is an example of buffering demand variability with time.
Examples (Non-Production)
Ballpoint Pens: Suppose a retailer sells ballpoint pens.
Demand is unpredictable (variable). But, since customers will go
elsewhere if they don't find the item in stock (who is going to backorder
a ballpoint pen?), the retailer cannot buffer this variability with time.
Likewise, because the instant delivery requirement of the customer rules
out a make-to-order environment, capacity cannot be used as a buffer.
This leaves only inventory. And indeed, this is precisely what the
retailer does by holding a stock of pens.
Ambulance and Fire Service: Consider emergency service providers.
Calls for service cannot be scheduled, so there is variability in demand.
However, because delays are disastrous, a time buffer is out of the question.
Since we cannot queue up calls for service, an inventory buffer is also
infeasible. Hence, the only option is a capacity buffer. And
indeed, ambulances and fire engines have very low utilization. The
excess capacity is a buffer against variability.
Organ Transplants: Demand for organs is variable, as is the supply. Since organs must be used quickly after they are removed from donors, an inventory buffer is impossible. Since the capacity is fixed by the supply of donated organs, a capacity buffer is also impossible. So, the only possibility is a time buffer. And indeed, waiting times for transplanted organs tend to be very long. The wait is a time buffer.