A repeated replenishment decision
A shop that orders frequently keeps little stock but pays the fixed ordering charge often. Ordering less frequently spreads that charge over more units but keeps more goods on the shelf. The economic order quantity model makes this tension explicit. It differs from a one-period uncertain-demand decision: here demand is deterministic and the same cycle repeats indefinitely.
Definition 8.7.1
(Economic order quantity model)
Demand occurs at a constant rate units per unit time. Whenever inventory reaches zero, a continuous quantity is replenished instantaneously. Each order costs a fixed , and holding one unit of inventory for one unit of time costs . There are no shortages, lead times, capacity limits, quantity discounts, deterioration or uncertainty, and fractional quantities are allowed.
The purchase cost per unit is constant, so its time-average contribution is the same for every and may be added back after optimization.
Derive the average cost
Proposition 8.7.2
(Cycle-average cost rate)
In the economic order quantity model, ignoring the constant purchase term, the long-run average cost per unit time of ordering each cycle is
Proof
(Inventory-time area of one cycle)
Immediately after an order the inventory is ; constant demand reduces it linearly to zero over a cycle of length . The inventory-time area is a triangle of area , so the holding cost per cycle is and average inventory over a cycle is . Adding the order charge and dividing by the cycle length gives . Every cycle is identical, so the long-run average equals this cycle average.
The two terms have the same units: cost per unit time. The average depends on the area under inventory, not the maximum inventory . The order size is a continuous model variable; if only whole units may be ordered, the feasible set and the final selection differ.
Common mistake
(Charging holding cost on the maximum inventory)
Writing the holding term as treats the shelf as full throughout the cycle. With , , , the incorrect objective is minimized at . Under the correct objective this order size costs per week, against the optimum found below. The holding term must use the time-average inventory .
The unique optimum
Theorem 8.7.3
(Economic order quantity)
Put . For every ,
Hence is the unique global minimizer of on , and the minimum cost rate is .
Proof
(Completing the square)
The definition of gives . Then , and for
Since and , the right side is nonnegative and vanishes only at . This proves the global optimum without relying on a stationary-point test alone.
Proof X-ray
Why the square closes the argument
The key move is the substitution , which rewrites both terms of over the common factor and exposes a perfect square. The hypothesis keeps that factor positive; makes the square term count and gives uniqueness; and make a genuine positive order size. Without the argument fails at the start: with , has infimum zero on but no minimizer. The identity also gives more than optimality. Dividing by ,
so the relative penalty depends only on the ratio and is the same for and .
At the optimum, ordering and holding costs are equal: . This equality follows from the optimal formula; it is not an independent assumption about all order sizes. If fixed ordering cost is multiplied by four while the other parameters stay fixed, doubles, not quadruples. Increasing makes orders more expensive, so the model orders less often and carries more stock. Increasing has the opposite effect. These comparative statements assume the other parameters and all model conditions remain unchanged.
Worked cycles
Worked example 8.7.4
(A weekly cycle at the optimum)
Let units per week, cost units per order and cost units per inventory unit per week. Then units. The cycle lasts week, with five orders per week and average inventory ten units. Ordering costs per week and holding costs per week; the variable total is per week, plus any constant purchase term .
Counterexample mode
The cheapest ordering rate alone is misleading
Using halves the order rate to per week, so ordering cost falls to per week. But average inventory rises to and holding cost to , giving total variable cost , greater than . One cost component can improve while the total worsens.
The ratio formula in the X-ray predicts how forgiving the cost curve is; the next example checks it.
Worked example 8.7.5
(A fifty percent error in the order size)
With the same parameters, take . Directly, . The ratio formula gives , and . The order size has the same ratio: . A fifty percent overshoot raises the cost by only about eight percent, and the penalty is symmetric in rather than in .
Worked example 8.7.6
(Whole units only)
Let , and , so and . If orders must be whole units, compare the two neighbors: and . The best integer order is , about percent above the continuous optimum. The comparison of both neighbors, not rounding, is what decides; see the warning below.
Boundaries and model changes
If a physical order must be an integer, and no other restrictions apply, compare and ; the continuous solution itself may not be feasible. This adjacency rule follows because the displayed square identity, or equivalently the derivative, shows decreases before and increases after it.
Common mistake
(Rounding the continuous optimizer to the nearest integer)
Nearest-integer rounding can pick the worse neighbor, because the cost depends on rather than on . With , , , the optimizer is , whose nearest integer is . Yet while , so is better. Both neighbors must be evaluated.
Replenishment delay, stochastic demand, lost sales, order-size discounts, or a finite planning horizon require a different model. The formula does not establish that those effects are negligible in a particular shop.
Exercises
Quick check
(For the worked parameters, calculate the variable cost rate at .)
Solution · Answer
per week.
Quick check
(If becomes and all other worked parameters remain fixed, find the new .)
Solution · Answer
Quadrupling doubles the optimal quantity from to , since .
Quick check
(With , identify the two adjacent positive integer order sizes and compare their costs.)
Solution · Answer
The continuous optimizer is , between one and two. and , so both are optimal among positive integer quantities.
Quick check
(Explain why adding a constant purchase cost per unit does not change the optimizer under the stated assumptions.)
Solution · Answer
Each cycle purchases units over time, so the long-run time-average purchase rate is , and purchase cost adds the same constant for all .