Lesson 34: The Production Model with Shortages Permitted¶
34.1 Learning Outcomes¶
By the end of this hour a student will be able to:
- Draw the four-phase inventory profile of the general production–backorder model.
- Derive \(Q^{*}\), \(I_{\max}\) and \(B^{*}\) for the combined model.
- Show that the general model reduces to each of Lessons 31–33 in the appropriate limit.
- Assemble the complete family of deterministic inventory models in one table.
- Select the right member of the family for a given situation.
34.2 The General Deterministic Model¶
This lesson combines the two extensions studied so far:
- Finite replenishment (Lesson 33): production at rate \(P>D\), not instantaneous delivery.
- Planned backorders (Lesson 32): shortages are permitted and cost \(C_s\) per unit per unit time.
The result is the most general deterministic EOQ-family model, with four parameters \((C_o, C_h, C_s, P)\) and two decision variables \((Q, I_{\max})\). Every earlier model is a special case.
The four-phase profile¶
stock
| /\ /\
Imax | / \ / \
| / \ / \
0 |---/----------\----------/----------\------- time
| / \ / \
-B | / \______/ \____
|<t1><--t2-->|<t3>|<-t4->|
|<---------- t ----------->|
| Phase | What is happening | Rate | Duration |
|---|---|---|---|
| \(t_1\) | Production on; backlog \(B\) being cleared | \(+(P-D)\) | \(B/(P-D)\) |
| \(t_2\) | Production on; stock building to \(I_{\max}\) | \(+(P-D)\) | \(I_{\max}/(P-D)\) |
| \(t_3\) | Production off; stock depleting to zero | \(-D\) | \(I_{\max}/D\) |
| \(t_4\) | Production off; backlog building to \(B\) | \(-D\) | \(B/D\) |
Production runs for \(t_1+t_2 = Q/P\) and the whole cycle is \(t=Q/D\).
The average levels¶
As in Lesson 32, both averages are triangle areas divided by the full cycle. Writing \(f=1-D/P\) and noting that the peak-to-trough range is \(Qf\) (the same total build-up as in Lesson 33):
Setting \(f=1\) recovers Lesson 32; setting \(B=0\) recovers Lesson 33.
34.3 Deriving the Optimum¶
Optimising over I_max¶
Result 1. $\(\boxed{\ I_{\max}^{*}=\frac{C_s}{C_h+C_s}\,Qf,\qquad B^{*}=\frac{C_h}{C_h+C_s}\,Qf\ }\tag{2.2}\)$
The same critical ratio \(\dfrac{C_s}{C_h+C_s}\) as in Lesson 32, now applied to \(Qf\) instead of \(Q\).
Optimising over Q¶
Substituting (2.2) into (2.1) reduces the holding-plus-shortage term to \(\dfrac{Qf}{2}\cdot\dfrac{C_hC_s}{C_h+C_s}\), giving an EOQ in disguise:
Result 2 (general production–backorder model). With $\(f=1-\frac{D}{P},\qquad k=\sqrt{\frac{C_h+C_s}{C_s}},\)$ $\(\boxed{\ Q^{*}=\sqrt{\frac{2C_oD}{C_hf}}\cdot k=\frac{Q^{*}_{\text{EOQ}}\cdot k}{\sqrt f}\ }\tag{2.3}\)$ $\(\boxed{\ I^{*}_{\max}=\frac{Q^{*}f}{k^{2}}=\frac{Q^{*}_{\text{EOQ}}\sqrt f}{k},\qquad B^{*}=Q^{*}f-I^{*}_{\max}\ }\tag{2.4}\)$ $\(\boxed{\ TRC^{*}=\frac{\sqrt{2C_oDC_h\,f}}{k}=\frac{TRC^{*}_{\text{EOQ}}\sqrt f}{k}\ }\tag{2.5}\)$
Both corrections multiply cleanly: \(\sqrt f\) from finite replenishment and \(1/k\) from backordering.
34.4 The Complete Family¶
| Model | \(P\) | \(C_s\) | \(Q^{*}\) | \(TRC^{*}\) | Lesson |
|---|---|---|---|---|---|
| Classical EOQ | \(\infty\) | \(\infty\) | \(\sqrt{\dfrac{2C_oD}{C_h}}\) | \(\sqrt{2C_oDC_h}\) | 31 |
| EOQ with backorders | \(\infty\) | finite | \(\sqrt{\dfrac{2C_oD}{C_h}}\,k\) | \(\dfrac{\sqrt{2C_oDC_h}}{k}\) | 32 |
| EPQ, no shortage | finite | \(\infty\) | \(\sqrt{\dfrac{2C_oD}{C_hf}}\) | \(\sqrt{2C_oDC_h f}\) | 33 |
| General | finite | finite | \(\sqrt{\dfrac{2C_oD}{C_hf}}\,k\) | \(\dfrac{\sqrt{2C_oDC_hf}}{k}\) | 34 |
with \(f=1-D/P\) and \(k=\sqrt{(C_h+C_s)/C_s}\).
One formula to remember. $\(Q^{*}=\sqrt{\frac{2C_oD}{C_h}}\times\underbrace{\frac{1}{\sqrt f}}_{\text{finite }P}\times\underbrace{k}_{\text{backorders}}, \qquad TRC^{*}=\sqrt{2C_oDC_h}\times\sqrt f\times\frac1k\)$ Both extensions increase \(Q\) and decrease \(TRC\), and they act independently.
34.5 Solved Problems¶
Problem 1 — The general model¶
\(D=10{,}000\) units/yr, \(P=25{,}000\) units/yr, \(C_o=₹150\), \(C_h=₹8\)/unit/yr, \(C_s=₹25\)/unit/yr. Find the complete policy and compare with all three earlier models.
Solution.
Step 1 — the two factors.
Step 2 — the batch.
Step 3 — the levels. First \(Q^*f = 908.30(0.6)=544.98\) — the peak-to-trough range.
Step 4 — the cost.
Step 5 — verification by components.
Again setup cost \(=\) holding \(+\) shortage.
Step 6 — the whole family, same data.
| Model | \(f\) | \(k\) | \(Q^{*}\) | Peak stock | \(TRC^{*}\) | Saving vs EOQ |
|---|---|---|---|---|---|---|
| EOQ (L31) | 1.0 | 1.0000 | 612.37 | 612.37 | ₹4898.98 | — |
| EOQ + backorders (L32) | 1.0 | 1.1489 | 703.56 | 533.00 | ₹4264.01 | 13.0 % |
| EPQ (L33) | 0.6 | 1.0000 | 790.57 | 474.34 | ₹3794.73 | 22.5 % |
| General (L34) | 0.6 | 1.1489 | 908.30 | 412.86 | ₹3302.89 | 32.6 % |
Interpretation. Each relaxation makes the batch larger and the cost lower, and the two effects compose multiplicatively: \(0.775\times0.870=0.674\), and indeed \(3302.89/4898.98=0.674\) ✓. Peak stock falls monotonically down the table — from \(612\) to \(413\) units, a \(33\,\%\) reduction in the warehouse space required, achieved while increasing the batch size by \(48\,\%\).
Problem 2 — Checking the limits¶
Verify that Result 2 reduces correctly in each of the three limiting cases, using Problem 1's data.
Solution.
(a) \(P\to\infty\) (instantaneous delivery). Then \(f\to1\):
— exactly Lesson 32's backorder model ✓.
(b) \(C_s\to\infty\) (no shortages tolerated). Then \(k\to1\) and \(I^*_{\max}\to Q^*f\), \(B^*\to0\):
— exactly Lesson 33's EPQ ✓.
(c) Both. \(f\to1\) and \(k\to1\):
— classical EOQ ✓.
Interpretation. Always test a general formula against its special cases. It costs two minutes, catches almost every algebraic slip, and is the technique by which the correctness of (2.3)–(2.5) can be confirmed without re-deriving them. A formula that fails a limit check is wrong, whatever the derivation looked like.
Problem 3 — Reading the cycle structure¶
For Problem 1's policy, find the four phase durations and state how long the firm is out of stock.
Solution. \(Q^*=908.30\), \(I^*_{\max}=412.86\), \(B^*=132.12\), \(P-D=15{,}000\)/yr.
| Phase | Formula | Value (yr) | Days |
|---|---|---|---|
| \(t_1\) clearing backlog | \(B/(P-D)=132.12/15000\) | 0.00881 | 3.2 |
| \(t_2\) building stock | \(I_{\max}/(P-D)=412.86/15000\) | 0.02752 | 10.0 |
| \(t_3\) depleting stock | \(I_{\max}/D=412.86/10000\) | 0.04129 | 15.1 |
| \(t_4\) building backlog | \(B/D=132.12/10000\) | 0.01321 | 4.8 |
| Total | 0.09083 | 33.2 |
Checks. Production time \(t_1+t_2=0.03632\) yr \(=Q^*/P=908.30/25000=0.03632\) ✓. Time in stock \(t_2+t_3 = 0.06881\); time short \(t_1+t_4=0.02202\).
which equals \(\dfrac{C_h}{C_h+C_s}=\dfrac{8}{33}=0.2424\) ✓ — the critical ratio again, exactly as in Lesson 32.
Interpretation. The firm operates a \(33\)-day cycle in which the machine runs for \(13.2\) days and customers wait for \(8.1\) days. Both are deliberate. Notice that the fraction of time short depends only on \(C_h\) and \(C_s\) — not on \(D\), \(C_o\) or \(P\). That makes it the single most robust number in the whole model and the natural thing to negotiate with a customer: "we will be out of stock about a quarter of the time; is that acceptable, and if not, what is it worth to you?"
Problem 4 — Selecting the right model¶
For each case, name the model and give the formula for \(Q^{*}\).
| Case | Situation | Model | \(Q^{*}\) |
|---|---|---|---|
| (a) | Buy screws from a supplier; delivery next day; no stock-outs allowed | EOQ | \(\sqrt{2C_oD/C_h}\) |
| (b) | Buy custom furniture; customers accept a 2-week wait | EOQ + backorders | \(\sqrt{2C_oD/C_h}\ k\) |
| (c) | Make brackets on a press at 400/day, use 150/day, no stock-outs | EPQ | \(\sqrt{2C_oD/(C_hf)}\) |
| (d) | Make machine tools at 8/week, demand 5/week, dealers accept backorders | General | \(\sqrt{2C_oD/(C_hf)}\ k\) |
| (e) | Make ice cream at 500 l/day, demand 300 l/day, cannot be stored over a week | None of these | — |
Case (e) explained. Perishability imposes a maximum cycle length, which none of these models contains. If the unconstrained \(Q^{*}\) gives \(t=Q^{*}/D\) longer than the shelf life \(T_{\max}\), the answer is simply \(Q=DT_{\max}\) — the constraint binds and the optimisation is irrelevant. Because the cost curve is flat (Lesson 31), the penalty for this forced choice is usually small, which is a comforting result: a binding practical constraint rarely costs much in an EOQ-family model.
Problem 5 — Sensitivity to C_s in the general model¶
For Problem 1's data, tabulate the general policy as \(C_s\) varies from \(₹2\) to \(\infty\).
Solution. \(f=0.6\) throughout, so \(Q^{*}_{\text{EPQ}}=790.57\) and \(TRC^{*}_{\text{EPQ}}=₹3794.73\) are the reference values.
| \(C_s\) | \(k\) | \(Q^{*}\) | \(Q^{*}f\) | \(I^{*}_{\max}\) | \(B^{*}\) | % of cycle short | \(TRC^{*}\) |
|---|---|---|---|---|---|---|---|
| 2 | 2.2361 | 1767.77 | 1060.66 | 212.13 | 848.53 | 80.0 % | ₹1697.06 |
| 8 | 1.4142 | 1118.03 | 670.82 | 335.41 | 335.41 | 50.0 % | ₹2683.28 |
| 25 | 1.1489 | 908.30 | 544.98 | 412.86 | 132.12 | 24.2 % | ₹3302.89 |
| 72 | 1.0541 | 833.33 | 500.00 | 450.00 | 50.00 | 10.0 % | ₹3600.00 |
| 792 | 1.0050 | 794.55 | 476.73 | 471.96 | 4.77 | 1.0 % | ₹3775.71 |
| \(\infty\) | 1.0000 | 790.57 | 474.34 | 474.34 | 0 | 0 % | ₹3794.73 |
Interpretation.
- \(I^{*}_{\max}\) rises and \(B^{*}\) falls as \(C_s\) increases, always summing to \(Q^{*}f\).
- The percentage of the cycle spent short is exactly \(C_h/(C_h+C_s)\) — check the \(C_s=8\) row: with \(C_h=C_s=8\) the system is short exactly half the time, regardless of everything else.
- As in Lesson 32, once \(C_s>10C_h\) (i.e. \(C_s>80\) here) the backorder refinement buys under \(5\,\%\) and is not worth the complexity. The general model earns its keep only when shortages are genuinely cheap.
34.6 Practice Problems¶
- \(D=6000\)/yr, \(P=15{,}000\)/yr, \(C_o=₹300\), \(C_h=₹10\), \(C_s=₹40\). Find \(Q^{*}\), \(I^{*}_{\max}\), \(B^{*}\) and \(TRC^{*}\). (Answer: \(f=0.6\), \(k=1.1180\); \(Q^{*}=866.03\), \(I^{*}_{\max}=415.69\), \(B^{*}=103.92\), \(TRC^{*}=₹4156.92\))
- Show that \(I^{*}_{\max}+B^{*}=Q^{*}f\) always, and explain what that quantity means physically.
- In Problem 1 of §4, what would \(C_s\) have to be for the firm to be out of stock only \(5\,\%\) of the time? (Answer: \(C_h/(C_h+C_s)=0.05\) gives \(C_s=₹152\))
- Prove that \(TRC^{*}\) for the general model is never greater than \(TRC^{*}\) for any of the three special cases with the same \(D\), \(C_o\) and \(C_h\).
- A firm using the general model finds \(Q^{*}=1000\), \(I^{*}_{\max}=480\) and \(B^{*}=120\). Deduce \(f\) and \(C_s/C_h\). (Answer: \(Q^*f = 600\) so \(f=0.6\), i.e. \(D/P=0.4\); \(B/(I_{\max}+B)=0.2=C_h/(C_h+C_s)\) so \(C_s=4C_h\))
34.7 Summary¶
- The general model has a four-phase cycle: clear backlog, build stock, deplete stock, build backlog.
- The peak-to-trough range is \(Qf\), and both averages divide by the full cycle: average stock \(=I_{\max}^2/(2Qf)\), average shortage \(=B^2/(2Qf)\).
- Critical ratio unchanged: \(I^{*}_{\max}=\dfrac{C_s}{C_h+C_s}Qf\), and the fraction of the cycle spent short is \(\dfrac{C_h}{C_h+C_s}\) — independent of \(D\), \(C_o\) and \(P\).
- \(\boxed{Q^{*}=\sqrt{\dfrac{2C_oD}{C_h}}\cdot\dfrac{k}{\sqrt f}}\) and \(\boxed{TRC^{*}=\sqrt{2C_oDC_h}\cdot\dfrac{\sqrt f}{k}}\) with \(f=1-D/P\), \(k=\sqrt{(C_h+C_s)/C_s}\).
- The two corrections compose multiplicatively and each one raises \(Q\) while lowering \(TRC\).
- Always check the limits \(P\to\infty\) and \(C_s\to\infty\): a general formula that does not collapse to its special cases is wrong.
- The backorder refinement is worth having only when \(C_s\) is within about an order of magnitude of \(C_h\); beyond that, use the simpler model.
- When a practical constraint (shelf life, storage space, minimum order) binds, take the constrained value — the flat cost curve makes the penalty small.
Next lesson: L35 — Quantity discounts and the consolidation of Unit III.
Practise now: unit3-queue-inventory-lab.html → General Model Lab: set \(P\) and \(C_s\) and watch
the four-phase profile, with each phase duration labelled and the cost breakdown updating live.