Skip to content

Lesson 35: Quantity Discounts and Consolidation of Unit III

35.1 Learning Outcomes

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

  1. Explain why a price break makes the total-cost curve discontinuous and the EOQ inapplicable.
  2. Apply the all-units discount algorithm correctly, including the feasibility test.
  3. Handle the case \(C_h=iC\), where the holding cost itself changes with the price.
  4. Solve a constrained multi-item inventory problem by Lagrange multipliers.
  5. Consolidate the whole of Unit III — queuing and inventory — into a single decision framework.

35.2 Why Discounts Break the EOQ

In Lessons 31–34 the unit price \(C\) was constant, so the purchase cost \(DC\) was a constant and could be dropped. With a quantity discount, \(C\) depends on \(Q\), so the purchase term must be kept:

\[TC(Q)=DC(Q)+\frac{C_oD}{Q}+\frac{C_hQ}{2}. \tag{1.1}\]

Two structural consequences:

  1. \(TC\) is discontinuous. At each price break it drops vertically, because the whole order suddenly becomes cheaper. Calculus cannot find the minimum of a discontinuous function.
  2. \(C_h\) usually changes too. With \(C_h=iC\), a lower price means a lower holding cost, so the EOQ for each price band is different: \(Q^{*}_{j}=\sqrt{2C_oD/(iC_j)}\).
  TC
    |\                             all-units discount
    | \    \                       (3 price bands)
    |  \    \       \
    |   \____\__     \____
    |         \  \___     \______
    |__________|________|__________ Q
              q1       q2

Each curve is valid only on its own interval; the answer is the lowest point of the visible envelope.

35.3 The All-Units Discount Algorithm

All-units discount. If \(Q\ge q_j\), then every unit costs \(C_j\). (Contrast the incremental discount, where only the units beyond \(q_j\) are cheaper — a different and less common scheme.)

The algorithm.

  1. Start with the cheapest price (the largest \(q_j\)) and work upwards in price.
  2. For price band \(j\), compute \(Q^{*}_{j}=\sqrt{\dfrac{2C_oD}{iC_j}}\).
  3. Feasibility test: is \(Q^{*}_{j}\ge q_j\) (within the band)?
    • Yes → this is a feasible EOQ. Compute \(TC\) at \(Q^{*}_{j}\). Stop testing cheaper bands; no lower price is attainable, so evaluate only this band and the bands above it at their break points.
    • No → the unconstrained optimum lies below the band's lower limit. Since \(TC\) is decreasing on the left of the EOQ, the best point within the band is its lower boundary \(q_j\). Evaluate \(TC(q_j)\).
  4. Compute \(TC\) for every candidate identified and choose the smallest.

The key insight. Within a band the cost curve is a convex EOQ curve. If the band's EOQ is infeasible (too small), the cheapest attainable point in the band is the break quantity \(q_j\) — because you are on the falling part of the curve. Never discard a band just because its EOQ is infeasible; the break point itself may still win, since the price is lower there.

35.4 Solved Problems

Problem 1 — A three-band all-units discount

Annual demand is \(2400\) units; ordering costs \(₹350\) per order; the carrying rate is \(25\,\%\) of unit price per year. The supplier quotes:

Order quantity Unit price
\(1\le Q<500\) ₹50
\(500\le Q<1000\) ₹48
\(Q\ge1000\) ₹45

Find the optimal order quantity.

Solution.

Step 1 — EOQ for each band (\(C_h=0.25C\)):

Band \(C\) \(C_h=0.25C\) \(Q^{*}_{j}=\sqrt{2(350)(2400)/C_h}\) Range Feasible?
3 (cheapest) ₹45 ₹11.25 \(\sqrt{1{,}680{,}000/11.25}=386.44\) \(Q\ge1000\) No (386 < 1000)
2 ₹48 ₹12.00 \(\sqrt{1{,}680{,}000/12}=374.17\) \(500\le Q<1000\) No (374 < 500)
1 ₹50 ₹12.50 \(\sqrt{1{,}680{,}000/12.5}=366.61\) \(Q<500\) Yes ✓

Step 2 — candidate quantities. Band 1's EOQ is feasible, so \(Q=366.61\) is a candidate. Bands 2 and 3 have infeasible (too small) EOQs, so their best points are their break quantities: \(Q=500\) and \(Q=1000\).

Step 3 — total cost of each candidate.

Candidate A: \(Q=366.61\) at ₹50

\[TC=2400(50)+\frac{350(2400)}{366.61}+\frac{12.50(366.61)}{2}=120{,}000+2291.29+2291.29=\mathbf{₹124{,}582.58}\]

Candidate B: \(Q=500\) at ₹48

\[TC=2400(48)+\frac{350(2400)}{500}+\frac{12(500)}{2}=115{,}200+1680.00+3000.00=\mathbf{₹119{,}880.00}\]

Candidate C: \(Q=1000\) at ₹45

\[TC=2400(45)+\frac{350(2400)}{1000}+\frac{11.25(1000)}{2}=108{,}000+840.00+5625.00=\mathbf{₹114{,}465.00}\]
\[\boxed{\text{Order }Q=1000\text{ units at ₹45; total cost ₹114{,}465 per year}}\]

Step 4 — the saving. Against the no-discount policy (Candidate A):

\[124{,}582.58-114{,}465.00=\mathbf{₹10{,}117.58\text{ per year}}\ (8.1\,\%)\]

Interpretation. Look at where the money is. Moving from A to C increases the inventory-related cost from \(₹4582.58\) to \(₹6465.00\) — the order is nearly three times larger, so holding cost more than doubles. But the purchase bill falls by \(₹12{,}000\), which swamps it. In a discount problem the purchase term dominates and the EOQ logic is almost a side-show. A \(10\,\%\) price cut on a ₹1.2 lakh purchase bill is worth far more than perfecting a ₹4600 inventory cost.

Note also the trap avoided: Band 3's EOQ was infeasible, and a careless solver would have discarded band 3 entirely — losing the best answer.

Problem 2 — When the discount is not worth taking

Same data as Problem 1, but the price breaks are now only ₹49.50 (at \(Q\ge500\)) and ₹49 (at \(Q\ge1000\)). Should the discount be taken?

Solution.

Band \(C\) \(C_h\) \(Q^{*}_{j}\) Feasible? Candidate \(Q\)
3 ₹49.00 ₹12.25 370.36 No 1000
2 ₹49.50 ₹12.375 368.49 No 500
1 ₹50.00 ₹12.50 366.61 Yes 366.61
\[TC_A=120{,}000+2291.29+2291.29=\mathbf{₹124{,}582.58}\]
\[TC_B=2400(49.50)+\frac{840{,}000}{500}+\frac{12.375(500)}{2}=118{,}800+1680+3093.75=\mathbf{₹123{,}573.75}\]
\[TC_C=2400(49)+840+\frac{12.25(1000)}{2}=117{,}600+840+6125=\mathbf{₹124{,}565.00}\]
\[\boxed{\text{Order }Q=500\text{ at ₹49.50; }TC=₹123{,}573.75}\]

Interpretation. The deepest discount is not the best. Taking the ₹49 price requires holding \(1000\) units, and the extra holding cost (\(₹6125\) versus \(₹3094\)) all but cancels the extra \(₹1200\) of price saving. The middle band wins by \(₹991\) over the deepest one.

This is the whole reason the algorithm exists. With shallow discounts, the answer is usually a middle band or no discount at all; with deep discounts (Problem 1), the deepest band usually wins. Only the arithmetic tells you which.

Problem 3 — Discounts with a fixed holding cost

Repeat Problem 1 but with the holding cost fixed at \(₹12\) per unit per year, independent of price (perhaps it is dominated by warehouse space, not capital).

Solution. Now there is a single EOQ for all bands:

\[Q^{*}=\sqrt{\frac{2(350)(2400)}{12}}=\sqrt{140{,}000}=374.17\text{ units}\]

This is feasible only in band 1 (\(Q<500\)). Candidates: \(374.17\) at ₹50, \(500\) at ₹48, \(1000\) at ₹45.

\[TC_A=120{,}000+\frac{840{,}000}{374.17}+\frac{12(374.17)}{2}=120{,}000+2245.00+2245.00=\mathbf{₹124{,}490.00}\]
\[TC_B=115{,}200+1680.00+3000.00=\mathbf{₹119{,}880.00}\]
\[TC_C=108{,}000+840.00+6000.00=\mathbf{₹114{,}840.00}\]
\[\boxed{Q=1000\text{ at ₹45; }TC=₹114{,}840}\]

Interpretation. The decision is unchanged, but the cost is \(₹375\) higher than in Problem 1 because the holding cost no longer falls with the price. The comparison isolates a real effect: when \(C_h=iC\), a discount is doubly attractive — it cuts both the purchase bill and the cost of holding the larger order. When \(C_h\) is fixed, only the first benefit applies. State which assumption you are making, because it changes the numbers even when it does not change the answer.

Problem 4 — A constrained multi-item problem

Three items are stocked in a warehouse with only \(600\) units of space. Independent EOQs exceed the space. Find the constrained policy.

Item \(D\) (units/yr) \(C_o\) (₹) \(C_h\) (₹/unit/yr)
1 5000 100 10
2 3000 150 5
3 2000 80 8

Solution.

Step 1 — unconstrained EOQs.

\[Q_1=\sqrt{\frac{2(100)(5000)}{10}}=316.23,\quad Q_2=\sqrt{\frac{2(150)(3000)}{5}}=424.26,\quad Q_3=\sqrt{\frac{2(80)(2000)}{8}}=200.00\]

Step 2 — check the constraint. Average space used is \(\sum Q_i/2\):

\[\frac{316.23+424.26+200.00}{2}=470.25\le 600\ \checkmark\]

The constraint is not binding, so the independent EOQs are optimal.

Step 3 — now suppose the space limit were \(400\) units. Then \(470.25 > 400\) and the constraint binds. Introduce a Lagrange multiplier \(\theta\ge0\) on the space constraint, which effectively raises every holding cost to \(C_{h_i}+2\theta\):

\[Q_i(\theta)=\sqrt{\frac{2C_{o_i}D_i}{C_{h_i}+2\theta}}\]

and choose \(\theta\) so that \(\sum Q_i/2 = 400\). Searching on \(\theta\):

\(\theta\) \(Q_1\) \(Q_2\) \(Q_3\) \(\sum Q_i/2\)
0.0 316.23 424.26 200.00 470.25
1.0 288.68 353.55 189.74 415.98
1.5 277.35 327.33 185.70 395.19
1.36 280.40 334.03 186.77 400.60
\[\boxed{\theta\approx1.36:\quad Q_1=280,\ Q_2=334,\ Q_3=187}\]

Interpretation. The multiplier \(\theta=1.36\) is the shadow price of warehouse space — exactly the concept from Lesson 7. It says that one more unit of space is worth about \(₹1.36\) per year, so the firm should rent extra space only if it costs less than that. The Lagrange multiplier turns a physical constraint into a price, which is the single most useful idea in constrained optimisation and reappears in Lessons 43–44.

Note that the reduction is not proportional: item 2 loses \(18.8\,\%\) of its batch, item 3 \(13.1\,\%\) and item 1 only \(10.9\,\%\). Items with low holding cost are squeezed hardest, because for them the added \(2\theta\) is a larger relative increase.

Problem 5 — Consolidation: choosing a model for Unit III

For each situation, identify the model from Lessons 21–35 and the key formula.

Situation Model Key result
One clerk, customers queue indefinitely \(M/M/1\) (L24) \(L_s=\rho/(1-\rho)\)
Four tellers, one shared line \(M/M/c\) (L27) Erlang-C, \(L_q=C(c,r)\rho/(1-\rho)\)
Car park with 40 bays, drivers leave if full \((M/M/c):(N)\) (L28) \(\lambda_{\text{eff}}=\lambda(1-P_N)\)
8 machines, 2 mechanics Finite source (L28) \(\lambda_n=(M-n)\lambda\)
How many tellers to hire Cost model (L29) \(TC=C_wL+C_sc\)
Buy 6000 bolts a year, no shortages EOQ (L31) \(Q^*=\sqrt{2C_oD/C_h}\)
Buy furniture, customers accept waiting EOQ + backorder (L32) \(S^*/Q^*=C_s/(C_h+C_s)\)
Make parts at 500/day, use 200/day EPQ (L33) \(Q^*=Q_{\text{EOQ}}/\sqrt f\)
Make parts, dealers accept backorders General (L34) \(Q^*=Q_{\text{EOQ}}k/\sqrt f\)
Supplier offers 5 % off above 1000 units Discount (L35) Test each band's EOQ for feasibility

The common structure of the whole unit. Every model in Lessons 21–35 balances a cost that falls with capacity or batch size against a cost that rises with it, and locates the minimum of a convex total. In queuing the trade-off is waiting cost against server cost; in inventory it is ordering cost against holding cost. The mathematics differs; the economics is identical.

Two further threads run through both halves:

  • Nonlinearity near the limit. Queues explode as \(\rho\to1\); batch sizes explode as \(P\to D\). Systems run near their capacity behave badly and unpredictably.
  • Flatness at the optimum. Both the queuing cost curve (L29) and the EOQ curve (L31) are flat near their minima, so approximate answers are cheap and precision in the inputs is rarely worth buying. This is what makes these models usable with imperfect data — and it is the most practically important fact in Unit III.

35.5 Practice Problems

  1. \(D=1000\)/yr, \(C_o=₹100\), \(i=20\,\%\). Prices: ₹20 for \(Q<200\); ₹18 for \(Q\ge200\). Find the optimal \(Q\). (Answer: band 2 EOQ \(=235.70\ge200\), feasible; \(TC=₹18{,}848.53\) versus \(₹20{,}632.46\) at band 1's EOQ of \(223.61\) — take the discount, \(Q=235.70\))
  2. In Problem 1 of §3, at what band-3 price would \(Q=1000\) stop being optimal? (Answer: equate \(2400C+840+\tfrac{0.25C(1000)}{2}\) to band 2's \(₹119{,}880\), i.e. \(2525C=119{,}040\), giving \(C=₹47.48\); any band-3 price above that loses to band 2)
  3. Show that with \(C_h=iC\) the EOQ for a cheaper band is always larger than for a dearer one, and explain why that makes a deep discount more likely to be feasible.
  4. Two items share \(500\) units of space; \(Q_1^*=400\), \(Q_2^*=300\) unconstrained. Is the constraint binding? (Answer: \(\sum Q_i/2=350\le500\) — not binding)
  5. A queue analysis recommends \(c=4\) servers and an inventory analysis recommends \(Q=800\). Both cost curves are flat near their optima. Write two sentences on what that implies for how the results should be presented to management.

35.6 Summary

  • A quantity discount makes \(C\) depend on \(Q\), so the purchase term \(DC(Q)\) must be kept and the total-cost curve becomes discontinuous.
  • All-units algorithm: compute each band's EOQ; if feasible it is a candidate, if infeasible (too small) the band's break quantity is the candidate. Evaluate every candidate's total cost.
  • Never discard a band because its EOQ is infeasible — the break point may still be optimal.
  • With \(C_h=iC\), each band has its own EOQ and a discount is doubly attractive. With fixed \(C_h\) there is one EOQ for all bands.
  • The deepest discount is not always best: extra holding cost can outweigh the price saving (Problem 2).
  • In discount problems the purchase term dominates. Optimising the batch size matters far less than securing the price.
  • Multi-item constraints are handled by a Lagrange multiplier that raises every holding cost to \(C_h+2\theta\); \(\theta\) is the shadow price of the scarce resource.
  • Unit III in one sentence: every model balances a decreasing cost against an increasing one; all behave badly near capacity and are forgivingly flat at the optimum.

Next lesson: L36 — Unit IV begins: introduction to non-linear programming, convex sets and convex functions.

Practise now: unit3-queue-inventory-lab.html → Discount Lab: enter price breaks and see the discontinuous cost curve with each band's EOQ marked feasible or infeasible.