Skip to content

Lesson 33: The Production (Finite Replenishment) Model without Shortages

33.1 Learning Outcomes

By the end of this hour a student will be able to:

  1. Explain how finite replenishment changes the inventory profile from sawtooth to trapezoid.
  2. Derive the maximum inventory \(I_{\max}=Q(1-D/P)\) from the geometry.
  3. Derive the Economic Production Quantity \(Q^{*}=\sqrt{\dfrac{2C_oD}{C_h(1-D/P)}}\).
  4. Interpret the inflation factor \(1/\sqrt{1-D/P}\) and the limits \(P\to\infty\) and \(P\to D\).
  5. Apply the model to batch-sizing and machine-loading decisions.

33.2 Why the EOQ Profile Is Wrong for a Factory

Classical EOQ assumes the whole order \(Q\) arrives at once. That is right for purchasing but wrong for manufacturing: a machine producing at \(P\) units per day delivers the batch gradually, and meanwhile the item is also being consumed at \(D\) units per day.

Economic Production Quantity (EPQ), also called the Economic Batch Quantity or the Economic Manufacturing Quantity. Same trade-off, different geometry.

Assumptions — those of Lesson 31, with assumption 4 replaced:

  • Production occurs at a finite, constant rate \(P\) units per unit time, with \(P > D\).
  • Consumption continues during production at rate \(D\).
  • No shortages permitted; all other assumptions unchanged.

The condition \(P>D\) is essential: if \(P\le D\), production can never get ahead of demand, no inventory ever accumulates and the machine must simply run continuously.

33.3 The Inventory Profile

 stock
      |      /\              /\
 Imax |    /   \           /   \
      |   /      \        /      \
      |  /         \     /         \       slope +(P-D) rising
      | /            \  /            \     slope  -D    falling
    0 |/_______________\/_______________\____ time
      |<-tp->|<- td ->|
      |<------- t ------>|
Phase Duration Net rate of change of stock
Production (produce and consume) \(t_p=Q/P\) \(+(P-D)\)
Depletion (consume only) \(t_d = t-t_p\) \(-D\)
Full cycle \(t=Q/D\) —

The profile is a trapezoid, not a sawtooth, and the peak is lower than \(Q\): some of the batch has already been consumed by the time the last unit is made.

Maximum inventory

Stock builds at rate \((P-D)\) for time \(t_p=Q/P\):

\[I_{\max}=(P-D)\frac{Q}{P}\]

Result 1. $\(\boxed{\ I_{\max}=Q\left(1-\frac{D}{P}\right)=Qf,\qquad f\equiv1-\frac{D}{P}\ }\tag{2.1}\)$ and the average inventory is \(\dfrac{I_{\max}}{2}=\dfrac{Qf}{2}\), not \(\dfrac{Q}{2}\).

The factor \(f=1-D/P\) is the fraction of the batch that is still on hand when production stops. It is the only structural difference from Lesson 31.

Check: \(P\to\infty\) gives \(f\to1\) and \(I_{\max}\to Q\) — the instantaneous-delivery sawtooth.

33.4 Deriving the Optimum

\[TRC(Q)=\underbrace{\frac{C_oD}{Q}}_{\text{setup}}+\underbrace{\frac{C_hQf}{2}}_{\text{holding}}\]

This is exactly the EOQ cost function with \(C_h\) replaced by \(C_hf\). So by the same calculus:

Result 2 (EPQ formulas). $\(\boxed{\ Q^{*}=\sqrt{\frac{2C_oD}{C_h\left(1-D/P\right)}}=\frac{Q^{*}_{\text{EOQ}}}{\sqrt{f}}\ }\tag{3.1}\)$ $\(\boxed{\ TRC^{*}=\sqrt{2C_oDC_h\left(1-\frac DP\right)}=TRC^{*}_{\text{EOQ}}\sqrt{f}\ }\tag{3.2}\)$ $\(I_{\max}=Q^{*}f,\qquad t=\frac{Q^{*}}{D},\qquad t_p=\frac{Q^{*}}{P},\qquad n=\frac{D}{Q^{*}}\)$

And, as always, at the optimum setup cost equals holding cost.

Reading the inflation factor

Everything is governed by \(1/\sqrt f = \left(1-D/P\right)^{-1/2}\):

\(D/P\) (machine utilisation) \(f\) \(Q^{*}/Q^{*}_{\text{EOQ}}\) \(TRC^{*}/TRC^{*}_{\text{EOQ}}\)
0 (instant supply) 1.00 1.000 1.000
0.25 0.75 1.155 0.866
0.50 0.50 1.414 0.707
0.75 0.25 2.000 0.500
0.90 0.10 3.162 0.316
0.99 0.01 10.000 0.100

The EPQ is always larger and always cheaper than the EOQ for the same data. Larger, because gradual delivery means a big batch never all sits in the store at once; cheaper, for the same reason — average stock is lower.

The limits are worth stating explicitly:

  • \(P\to\infty\): \(f\to1\), and (3.1)–(3.2) become the EOQ formulas exactly.
  • \(P\to D\): \(f\to0\), so \(Q^{*}\to\infty\) and \(TRC^{*}\to0\). Interpretation: a machine that can barely keep up must run continuously, so there are no setups and no accumulated stock. The formula's advice ("produce infinitely large batches") really means "never stop".

33.5 Solved Problems

Problem 1 — The standard EPQ calculation

A firm needs \(10{,}000\) units a year of a part it makes itself. The machine produces \(25{,}000\) units a year. Setup costs \(₹150\); holding costs \(₹8\) per unit per year. Find \(Q^*\), \(I_{\max}\), \(TRC^*\), the cycle time, the production time per cycle and the number of runs per year.

Solution.

\[f=1-\frac{D}{P}=1-\frac{10000}{25000}=1-0.4=0.6\]
\[Q^{*}=\sqrt{\frac{2(150)(10000)}{8(0.6)}}=\sqrt{\frac{3{,}000{,}000}{4.8}}=\sqrt{625000}=\mathbf{790.57\text{ units}}\]
\[I_{\max}=790.57(0.6)=\mathbf{474.34\text{ units}}\]
\[TRC^{*}=\sqrt{2(150)(10000)(8)(0.6)}=\sqrt{14{,}400{,}000}=\mathbf{₹3794.73\text{ per year}}\]

Check: setup \(=\dfrac{150(10000)}{790.57}=₹1897.37\); holding \(=\dfrac{8(474.34)}{2}=₹1897.37\) ✓ equal.

\[t=\frac{790.57}{10000}=0.0791\text{ yr}=\mathbf{28.9\text{ days}},\qquad t_p=\frac{790.57}{25000}=0.0316\text{ yr}=\mathbf{11.5\text{ days}}\]
\[n=\frac{10000}{790.57}=\mathbf{12.65\text{ runs per year}}\]

Comparison with EOQ (Lesson 31, same data): \(Q^{*}_{\text{EOQ}}=612.37\), \(TRC^{*}=₹4898.98\).

EOQ (bought in) EPQ (made in-house)
Batch size 612.37 790.57 (\(\times1.291\))
Peak stock 612.37 474.34
Average stock 306.19 237.17
Relevant cost ₹4898.98 ₹3794.73 (\(\times0.775\))

Interpretation. The batch is \(29\,\%\) larger and yet the peak stock is \(23\,\%\) smaller, so the cost falls by \(23\,\%\). Both effects come from the same fact: \(40\,\%\) of each batch is consumed before the run finishes. The machine runs \(11.5\) days and stands idle (for this part) for \(17.4\) days in every \(28.9\)-day cycle — which is the capacity available for other products.

Problem 2 — A high-utilisation machine

A plant needs \(9000\) units a year; the machine can make \(12{,}000\) a year. Setup costs \(₹400\), holding \(₹6\) per unit per year. Find the policy, and compare it with the EOQ answer.

Solution.

\[f=1-\frac{9000}{12000}=1-0.75=0.25\]
\[Q^{*}=\sqrt{\frac{2(400)(9000)}{6(0.25)}}=\sqrt{\frac{7{,}200{,}000}{1.5}}=\sqrt{4{,}800{,}000}=\mathbf{2190.89\text{ units}}\]
\[I_{\max}=2190.89(0.25)=\mathbf{547.72\text{ units}}\]
\[TRC^{*}=\sqrt{2(400)(9000)(6)(0.25)}=\sqrt{10{,}800{,}000}=\mathbf{₹3286.34\text{ per year}}\]
\[t=\frac{2190.89}{9000}=0.2434\text{ yr}=\mathbf{88.9\text{ days}},\qquad t_p=\frac{2190.89}{12000}=0.1826\text{ yr}=\mathbf{66.6\text{ days}}\]
\[n=\mathbf{4.11\text{ runs per year}}\]

Against EOQ: \(Q^{*}_{\text{EOQ}}=\sqrt{2(400)(9000)/6}=1095.45\) and \(TRC^{*}_{\text{EOQ}}=₹6572.67\).

\[\frac{Q^{*}_{\text{EPQ}}}{Q^{*}_{\text{EOQ}}}=\frac{1}{\sqrt{0.25}}=2.000,\qquad \frac{TRC^{*}_{\text{EPQ}}}{TRC^{*}_{\text{EOQ}}}=\sqrt{0.25}=0.500\]

Interpretation. At \(75\,\%\) machine utilisation the batch doubles and the cost halves relative to the purchased case. Note also that the machine is producing this part for \(66.6\) of every \(88.9\) days — it is nearly dedicated to this product. Using the EOQ formula here would have recommended batches of \(1095\), less than half the right size, and overstated the cost by a factor of two. Ignoring finite replenishment is a large error, not a small one.

Problem 3 — Sensitivity to the production rate

For Problem 1's data (\(D=10{,}000\), \(C_o=₹150\), \(C_h=₹8\)), tabulate the policy as \(P\) varies.

Solution.

\(P\) \(D/P\) \(f\) \(Q^{*}\) \(I_{\max}\) \(TRC^{*}\)
\(\infty\) 0.000 1.000 612.37 612.37 ₹4898.98
100 000 0.100 0.900 645.50 580.95 ₹4647.58
40 000 0.250 0.750 707.11 530.33 ₹4242.64
25 000 0.400 0.600 790.57 474.34 ₹3794.73
20 000 0.500 0.500 866.03 433.01 ₹3464.10
12 500 0.800 0.200 1369.31 273.86 ₹2190.89
10 500 0.952 0.0476 2806.24 133.63 ₹1069.04

Interpretation.

  • \(Q^{*}\) rises steeply and \(I_{\max}\) falls as \(P\) approaches \(D\). Batch size and peak stock move in opposite directions — a genuinely counter-intuitive result that follows directly from \(I_{\max}=Q f\) with \(Q\propto f^{-1/2}\), so \(I_{\max}\propto \sqrt f\).
  • The last row is a warning: the formula recommends a batch of \(2806\) — nearly a third of annual demand — because the machine has almost no spare capacity. That is mathematically right and practically fragile: any breakdown becomes a stock-out, and the model has no concept of risk. At \(D/P>0.8\), treat the recommendation as an indication and add a capacity or risk constraint.

Problem 4 — Choosing between making and buying

A firm can buy an item for \(₹52\) per unit (\(C_o=₹150\) for a purchase order) or make it at a variable cost of \(₹48\) per unit on a machine running at \(25{,}000\) units a year with a setup cost of \(₹450\). Demand is \(10{,}000\) a year and \(C_h=₹8\) per unit per year in both cases. Which is cheaper?

Solution. Here the purchase/production cost is decision-relevant, because it differs between the options — unlike in Lessons 31–32.

Option A — buy. \(Q^{*}=612.37\), \(TRC^{*}=₹4898.98\) (Lesson 31).

\[TC_A = 10000(52)+4898.98 = 520000+4898.98=\mathbf{₹524{,}898.98}\]

Option B — make. \(f=0.6\), \(C_o=450\):

\[Q^{*}=\sqrt{\frac{2(450)(10000)}{8(0.6)}}=\sqrt{1{,}875{,}000}=1369.31\text{ units}\]
\[TRC^{*}=\sqrt{2(450)(10000)(8)(0.6)}=\sqrt{43{,}200{,}000}=₹6572.67\]
\[TC_B = 10000(48)+6572.67=480000+6572.67=\mathbf{₹486{,}572.67}\]
\[\boxed{\text{MAKE — saving }₹38{,}326.31\text{ per year}}\]

Interpretation. The decision is driven almost entirely by the \(₹4\) per unit difference in variable cost (\(₹40{,}000\) a year); the inventory costs differ by only \(₹1674\) and move the other way. When comparing options with different unit costs, the purchase term dominates and must be included — but notice that the higher setup cost of making did make the optimal batch more than twice as large, which is exactly the kind of operational consequence a purely accounting comparison would miss.

A complete analysis would also weigh the capacity consumed (\(t_p/t = D/P = 40\,\%\) of the machine's time), which is unavailable to other products and has an opportunity cost this model does not capture.

Problem 5 — Multi-product machine loading

A machine makes three products. Using the EPQ for each, check whether the machine has enough capacity.

Product \(D\) (units/yr) \(P\) (units/yr) \(C_o\) (₹) \(C_h\) (₹/unit/yr)
X 10 000 25 000 150 8
Y 6 000 30 000 200 5
Z 4 000 20 000 100 6

Solution.

Product \(f\) \(Q^{*}\) \(n=D/Q^{*}\) \(t_p\) per run (yr) Run time per year \(=D/P\)
X 0.60 790.57 12.65 0.03162 0.4000
Y 0.80 774.60 7.75 0.02582 0.2000
Z 0.80 408.25 9.80 0.02041 0.2000
30.19 setups/yr 0.8000

Capacity check. Total production time is \(\sum D/P = 0.40+0.20+0.20=\mathbf{0.80}\), i.e. \(80\,\%\) of the year. That leaves \(20\,\%\) — about \(50\) working days — for setups, maintenance and slack.

If each setup takes \(4\) hours on a \(2000\)-hour year, \(30.19\) setups consume \(30.19\times4/2000=0.0604\), i.e. a further \(6.0\,\%\). Total load \(=86.0\,\%\).

\[\boxed{\text{Feasible, at }86.0\,\%\text{ utilisation — but with little margin}}\]

Interpretation. Two points that the single-product model cannot see.

  1. The EPQs were computed independently, yet the machine can make only one product at a time. Nothing guarantees the three cycles can be scheduled without collision. A proper treatment uses a common cycle time for all products, which is the subject of the multi-item extension.
  2. At \(86\,\%\) utilisation, Lesson 24's lesson applies: this machine is on the steep part of the congestion curve. A breakdown or a demand increase will cause queueing of jobs and missed schedules. A capacity check is not optional after an EPQ calculation — the formula happily recommends batch sizes for a machine that does not have the hours.

33.6 Practice Problems

  1. \(D=4800\)/yr, \(P=8000\)/yr, \(C_o=₹250\), \(C_h=₹12\). Find \(Q^*\), \(I_{\max}\), \(TRC^*\), \(t\) and \(t_p\). (Answer: \(707.11\), \(282.84\), \(₹3394.11\), \(0.1473\) yr, \(0.0884\) yr)
  2. Show that \(I_{\max}\propto\sqrt f\), and hence that peak stock falls as the machine gets busier.
  3. A firm computes EOQ instead of EPQ for a part with \(D/P=0.6\). By what percentage is its batch size wrong, and by how much does its cost exceed the true optimum? (Answer: batch \(36.75\,\%\) too small; using the sensitivity formula of Lesson 31 with \(k=\sqrt{0.4}=0.6325\) gives \(TRC/TRC^*=1.1068\), i.e. \(10.7\,\%\) too dear)
  4. At what value of \(D/P\) does the EPQ batch become exactly twice the EOQ batch? (Answer: \(f=0.25\), i.e. \(D/P=0.75\))
  5. In Problem 5, demand for X rises to \(15{,}000\)/yr. Recompute the capacity check and comment. (Answer: \(D/P\) for X becomes \(0.60\); total production load \(1.00\) — the machine is fully saturated with no time for setups. Infeasible.)

33.7 Summary

  • Finite replenishment turns the sawtooth into a trapezoid: production and consumption overlap.
  • The whole model is EOQ with \(C_h\) replaced by \(C_hf\), where \(\boxed{f=1-D/P}\).
  • \(\boxed{I_{\max}=Qf}\) and average stock \(=Qf/2\) — not \(Q/2\).
  • \(\boxed{Q^{*}=\dfrac{Q^{*}_{\text{EOQ}}}{\sqrt f}}\) and \(\boxed{TRC^{*}=TRC^{*}_{\text{EOQ}}\sqrt f}\). Bigger batches, lower cost, always.
  • \(I_{\max}\propto\sqrt f\), so batch size and peak stock move in opposite directions as \(P\) falls towards \(D\).
  • \(P\to\infty\) recovers EOQ exactly; \(P\to D\) says "never stop the machine".
  • Ignoring finite replenishment is a large error: at \(D/P=0.75\) it halves the batch and doubles the reported cost.
  • Always finish with a capacity check: \(\sum D_i/P_i\) plus setup time must be comfortably below 1.

Next lesson: L34 — The production model with shortages permitted, combining Lessons 32 and 33.

Practise now: unit3-queue-inventory-lab.html → EPQ Lab: drag \(P\) and watch the trapezoid flatten, \(Q^{*}\) grow and \(I_{\max}\) shrink at the same time.