Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
The investigation of dynamic inventory models has largely assumed price to be constant and concerned itself with the production decision only. Notable exceptions are
[3] and [4]. It would seem worthwhile in many situations to consider the total effect
on profits of price-production decisions and make the two decisions simultaneously.
This paper considers the problem of a monopolist who must set both decisions in each
of T periods. The inventory work of Wagner and Whitin [5] and a paper based on their
work by Eppen, Gould and Pashigian [1] are extended to the price-production case.
The major results are in obtaining planning horizons similar to those in [1], [5], and [6]
for the price-production atuaiion and in limiting the calculations required for the pricing decision.
Fonnulation
Confflder a deterministic demand function, nonincreasing in price and different in
each of T periods. No backorders are allowed, and profit maximization is the objective.
We assume there exist sets P, D, and X where:
P = {p: p is an admissble price}
I) jd: d is an admissible demand}
X = jz: z is an admissible production level}
P and X contain ail decisions that are being considered (pi and Xt,i = 1, , T),
and X is such that no lapadty constraint exists. There is a demand function as specified
above which is denoted:
d = *!&><)> where we assume *(?) is such that pr4>t(pi) takes on its maximum
value for a finite value of pi in all periods.
The immediate costs in each period, L, (), are given by equation 1.
(1)
where:
/
it-i
Kt
Vt
i{xt)
L,(I, X,, p,) = it-il + h{Xi)Kt + F,z, = inventory level at the b^inning of period t.
= invraitory charge to carry one unit from t 1 to t.
= set-up charge in period t.
variable cost per unit in period t.
= 0 if I, = 0,
= 1 if z, > 0.
pMVt)
748
JOSEPH THOMAS
Reference [5j gives a forward algorithm based on four theorems with no price decision.
The price production analogs of these theorems are stated below without proof since
the proofs follow directly from those in [5] and [6].
L E M M A 1. There exists an optimal program such that Ii-Xi
= 0 / o r aU t.
Based on these, we can give the following forward algorithm which optimally chooses
prices and production periods, which can be translated into production quantities. Let
Pit = (Pi, , Pi),
with P i , , p . P ,
Then
(2)
= I, ,T
and F ( 0 ) = 0.
P l a n n i n g Horizons
Let S{tx, fe) = {i U < e U , 2, , fe}: F , -h J^^
(3)
> A lon^r version of the p^>er contuning these and other proofs and numerical examples is
available from the author.
749
PBOOF. We know (3) takes on its rnftinmiim for a finite pi since pf<^i(pt) does so.
Call the TnaTiTnigiTig price for (3) pfo.t. Assume some other price pt is optimal, and it is
sufficient to show tiiat changing to pft.j would never decrease the profit. Such a change
implies a net increase of:
(4)
since no set-up has been changed. But (4) is nonnegative by the assumption of (3),
proving the theorem.
DBCTJSSION. The implication is that prices between set-ups may be chosen one at a
time. Also, if (e.g.) the period t problem xmes a set-up inf 2,in<-|-lwe may evaluate
a set-up iat 2 using the t 2,t 1 and t period prices chosen in the period t problem. Further, since the proof does not depend on f', we can a priori evaluate all possible
optimal prices with at most T{T -\- l)/2 one dimensional maximizations. These calculations are relatively easy, as can be seen with the following linear demand example.
Suppose:
(5)
il>t(pt) = at btPt,
bt>0.
Define: a-, = (p V,) {at feip<)-Then let p,* be the price which maximizes that
value, and it can be demonstrate that:
(6)
If period t is satisfied from f < t. Theorem 2 indicates an optimal price, pfo,,. From
(3) we see that:
(7)
(8)
That is, the optimal price, pfo,,, equals p* plus one half the difference in variable
cost of satisfying t from f or from t itself, jmd this is easily computed. If a finite number
of prices and a more complex demand function are under consideration, however, the
optimization in Theorem 2 may involve enumeration. In that case it is useful to restrict
the number of possible optimal prices, and the two results below (see [2] for proofs)
do tioB.
LsuMA 6. Ifm(t) ^ t,no price lower than pt* is ever optimal.
LEHUA7. Ifm(t) t, no price, Pt, higher than Pi*, su(^ that the one period contribution, Xt, using pt is smaUer by ai least Kt Uwn the coiUribution using pt* is ever optimal.
Algorithm
The i ^ r i t h m , whidi uses the above results and is similar to the algorithm in [1], is
given below.
Step 1. Eecord m(t), b^inning with t = 1.
Step 2. a. R^rtsrict poi^ble set-ups to S(j(.t 1), 0b. If P is finite, m{t) = t and the use of Theorem 2 for all poasible set-ups
would be computationally difficult, restrict the posable prices in accordance with
Lemmas 6 and 7.
Step 3. Uang the restricted deddon space, find an optimal solution to the {-period
750
JOSEPH THOMAS
problem using (2) and Theorem 2; record optimal prices in conjunetion witii eaeh possible last %t-up.
Step 4.1{j(t) = m(t),j(t) 1 is a plamung horizon, production dmsionstoi(0
1 and pricing dedaons to j(t) are optimal in the T-period pn^ram, and they may be
recorded as such. (Step 1 guarantees that they are not reconsidered.)
Step 5. Record F(t) andi(0. (Prices are recorded in 3.)
Step 6. Repeat 1 to 6 for < + 1, , T. F{t), periods where set-ups occur, and optimal prices are avdlable, and production quantities can be computed.
References
1. EPFBN, G. D . , Gotru), F. J., AND FASHiaiAN, B. P., "Extemions of the Planning Horison
Tteorem in the Djrnsmio Economic Lot Siie Model," Management Science, Vol. 15, No. 5
(January 19%), pp. 2^-^277.
2. THOMAa, L. J., "Simultaneous Price-Production Decisions with Deterministic and Random
Demand," impublished Ph.D. dissertation, Yale University, 1968.
3. WAQNBK, H . M . , "A Postscript to "Dynamic Problems of the Firm,''' Naval Research Loffisticg
Quarterly, Vol. 7, No. 1 (1960), pp. 7-12.
4.
, AND WHITIN, T . M . , "Dynamic Problems in the Theory of the Firm," Naval Research
Logistics QuarUrly, Vol. 5, No. 1 (1958), pp. 53-74.
6.
, and
, "Dynamic Version of the Economic Lot Size Formula," Management Science,
Vol. 5, No. 1 (October 19ffi), pp. 89-96.
6. ZABEL, E . , "Some Generalizations of an Inventory Planning Horizon Theorem," Management
Science, Vol. 10, No. 3 (April 1964), pp. 465-471.