Inventory Models: EOQ, Shortages, Production and Newsvendor
CSIR NET elementary inventory models: EOQ, planned shortages, finite production rate, quantity discounts, reorder point and the single-period newsvendor.
- 17 min
On this page · 15 sections
- Prerequisites
- The Hook: Order Often or Order Big?
- Notation and Assumptions
- Model 1: The Basic EOQ
- Model 2: EOQ with Planned Shortages (Backorders)
- Model 3: Finite Production Rate (EPQ)
- Model 4: Quantity Discounts
- Model 5: The Single-Period (Newsvendor) Model
- Memory Tricks
- Don’t Confuse
- Common Mistakes
- Practice
- Quick Revision Table
- Key Takeaways
- What’s Next
Prerequisites
- Continuity, Differentiability and Bounded Variation: every model here minimises a smooth convex cost function of one or two variables by setting derivatives to zero.
- Metric Spaces and Several Variables: the shortage model has two decision variables and needs the two-variable second-derivative test.
- Percentages, Ratios and Averages (Part A): holding costs are often quoted as a percentage of unit price per year.
- The chapter map is on Reliability and Operations Research.
The Hook: Order Often or Order Big?
Notation and Assumptions
| Symbol | Meaning | Units |
|---|---|---|
| \( D \) | demand rate | units per year |
| \( A \) | ordering (setup) cost per order | ₹ per order |
| \( h \) | holding cost | ₹ per unit per year |
| \( s \) | shortage (backorder) cost | ₹ per unit short per year |
| \( P \) | production rate (\( P > D \)) | units per year |
| \( c \) | purchase cost per unit | ₹ per unit |
| \( Q \) | order (lot) size | units |
| \( L \) | lead time | years (or days) |
The total relevant cost per year includes only the costs that depend on the decision. In the basic models the purchase cost \( cD \) does not depend on \( Q \) and is left out; it comes back when prices depend on \( Q \) (discounts).
Model 1: The Basic EOQ
Assumptions: demand deterministic and constant at rate \( D \); instantaneous replenishment; no shortages; zero (or known, constant) lead time; costs constant.
Inventory falls linearly from \( Q \) to 0 in each cycle of length \( t = \frac{Q}{D} \), then jumps back to \( Q \) (a saw-tooth). Average inventory is \( \frac{Q}{2} \).
$$TC(Q) = \underbrace{\frac{AD}{Q}}_{\text{ordering}} + \underbrace{\frac{hQ}{2}}_{\text{holding}}.$$Setting \( TC'(Q) = -\frac{AD}{Q^2} + \frac{h}{2} = 0 \) (and \( TC'' = \frac{2AD}{Q^3} > 0 \)):
$$\boxed{Q^* = \sqrt{\frac{2AD}{h}}, \qquad TC^* = \sqrt{2ADh} = hQ^*, \qquad t^* = \frac{Q^*}{D}, \qquad N^* = \frac{D}{Q^*}.}$$For the hook: \( Q^* = \sqrt{\frac{2 \times 100 \times 1200}{6}} = \sqrt{40000} = 200 \), \( TC^* = \sqrt{2 \times 100 \times 1200 \times 6} = 1200 \) ₹ per year, 6 orders a year, one every 2 months.
Lead time and the reorder point
If an order takes \( L \) to arrive, place it when stock falls to the reorder point \( R = DL \) (demand during lead time).
If \( L \) is longer than a cycle, several orders are outstanding at once: the on-hand stock at reordering is \( D\,(L - k t^*) \) where \( k = \lfloor L/t^* \rfloor \), while the inventory position (on hand plus on order) is still \( DL \).
Example: 20 units per day, \( Q = 100 \) (cycle 5 days), lead time 12 days. \( DL = 240 \); two orders (200 units) are always in the pipeline, so reorder when on-hand stock reaches \( 20 \times (12 - 10) = 40 \).
Model 2: EOQ with Planned Shortages (Backorders)
Now shortages are allowed and fully backordered: demand during a stock-out is met when the next order arrives, at cost \( s \) per unit per year short. Let \( S \) be the maximum inventory (stock right after the backorders are filled), so the maximum shortage is \( Q - S \).
Over a cycle of length \( \frac{Q}{D} \): stock is positive for a fraction \( \frac{S}{Q} \) of the time with average \( \frac{S}{2} \), and negative for a fraction \( \frac{Q - S}{Q} \) with average shortage \( \frac{Q - S}{2} \). So
$$TC(Q, S) = \frac{AD}{Q} + \frac{hS^2}{2Q} + \frac{s(Q - S)^2}{2Q}.$$Setting both partial derivatives to zero:
$$\boxed{Q^* = \sqrt{\frac{2AD}{h}}\sqrt{\frac{h + s}{s}}, \qquad S^* = \sqrt{\frac{2AD}{h}}\sqrt{\frac{s}{h + s}} = \frac{s}{h + s}\,Q^*, \qquad TC^* = \sqrt{2ADh}\,\sqrt{\frac{s}{h + s}}.}$$- Allowing shortages increases the lot size and decreases the cost (by the factor \( \sqrt{\frac{s}{h + s}} < 1 \)).
- As \( s \to \infty \) (shortages prohibitive), the factors tend to 1 and Model 1 is recovered.
- The fraction of each cycle in stock-out is \( \frac{Q^* - S^*}{Q^*} = \frac{h}{h + s} \).
Model 3: Finite Production Rate (EPQ)
Instead of arriving all at once, the lot is produced at rate \( P > D \) while demand continues at rate \( D \). During production, stock builds at rate \( P - D \) for a time \( \frac{Q}{P} \), so the maximum inventory is
$$I_{\max} = (P - D)\frac{Q}{P} = Q\left(1 - \frac{D}{P}\right),$$and the average is half of that. Hence
$$TC(Q) = \frac{AD}{Q} + \frac{hQ}{2}\left(1 - \frac{D}{P}\right), \qquad \boxed{Q^* = \sqrt{\frac{2AD}{h\,(1 - D/P)}}, \qquad TC^* = \sqrt{2ADh\left(1 - \frac{D}{P}\right)}.}$$The model is Model 1 with \( h \) replaced by \( h(1 - \frac{D}{P}) \). As \( P \to \infty \) it becomes Model 1; as \( P \to D \) the lot size grows without bound (production runs continuously and no stock accumulates).
| Cycle quantity | Formula |
|---|---|
| Production run time | \( \frac{Q}{P} \) |
| Cycle length | \( \frac{Q}{D} \) |
| Maximum inventory | \( Q(1 - \frac{D}{P}) \) |
| Fraction of cycle producing | \( \frac{D}{P} \) |
Production with shortages. Combining both corrections, the optimal lot is \( Q^* = \sqrt{\frac{2AD}{h(1 - D/P)}}\sqrt{\frac{h + s}{s}} \) and \( TC^* = \sqrt{2ADh(1 - \frac{D}{P})}\sqrt{\frac{s}{h + s}} \): each modification contributes its own factor.
Model 4: Quantity Discounts
Suppose the unit price depends on the lot size (all-units discount): price \( c_j \) applies to the whole order when \( Q \) lies in the \( j \)-th price range, and the holding cost is a fraction \( i \) of the price, \( h_j = i c_j \). Now the purchase cost matters:
$$TC_j(Q) = Dc_j + \frac{AD}{Q} + \frac{i c_j Q}{2}.$$Procedure.
- For each price \( c_j \), compute \( Q_j^* = \sqrt{\frac{2AD}{ic_j}} \).
- If \( Q_j^* \) lies in its own range, it is a candidate. If it is below the range, the candidate is the range’s lower break point (the convex cost increases away from \( Q_j^* \)). If it is above the range, that price is dominated (a lower price serves that quantity).
- Evaluate \( TC_j \) at every candidate, including the purchase cost, and choose the smallest.
Model 5: The Single-Period (Newsvendor) Model
Some items are ordered once for one selling season (newspapers, seasonal goods, perishable reagents). Demand \( X \) is random with distribution function \( F \). Let
- \( c_u \) = underage cost per unit (profit lost when demand exceeds stock: price minus cost),
- \( c_o \) = overage cost per unit (loss on each unsold unit: cost minus salvage value).
Marginal argument. Stocking the \( Q \)-th unit gains \( c_u \) with probability \( P(X \ge Q) \) and loses \( c_o \) with probability \( P(X < Q) \). It is worth stocking while \( c_u P(X \ge Q) \ge c_o P(X < Q) \). Hence
$$\boxed{F(Q^*) = \frac{c_u}{c_u + c_o} \quad \text{(continuous demand)}}$$and for discrete demand, \( Q^* \) is the smallest \( Q \) with \( F(Q) \ge \frac{c_u}{c_u + c_o} \). The ratio \( \frac{c_u}{c_u + c_o} \) is the critical ratio (the optimal probability of not running out).
graph TD
A{"Demand known and steady?"} -- "Yes" --> B{"Replenishment instantaneous?"}
A -- "No, one season" --> N["Newsvendor: F(Q) = cu / (cu + co)"]
B -- "Yes" --> C{"Shortages allowed?"}
B -- "No, rate P" --> E["EPQ: h becomes h(1 - D/P)"]
C -- "No" --> D["EOQ: Q = sqrt(2AD/h)"]
C -- "Yes" --> F["Backorders: extra factor sqrt((h+s)/s)"]
D --> G{"Price depends on Q?"}
G -- "Yes" --> H["Discounts: compare total cost incl. purchase at candidates"]Memory Tricks
Don’t Confuse
| This | vs | That |
|---|---|---|
| EOQ \( \sqrt{2AD/h} \): order size | \( \sqrt{2ADh} \): minimum annual cost | |
| Maximum inventory \( S^* \) in the shortage model | order quantity \( Q^* \) (larger by the backorder \( Q^* - S^* \)) | |
| EPQ maximum inventory \( Q(1 - \frac{D}{P}) \) | EPQ lot size \( Q \) | |
| On-hand stock at reorder | inventory position (on hand + on order) \( = DL \) | |
| Discount: compare total cost with purchase cost | basic EOQ: purchase cost irrelevant | |
| Newsvendor underage \( c_u \) = price − cost | overage \( c_o \) = cost − salvage |
Common Mistakes
Practice
All questions below are original practice questions written for this page. They are not from any CSIR NET paper. In Level 3, as in Part C, one or more options may be correct.
Level 1: Foundation
- Write the EOQ formula and name each symbol.
- At the EOQ, how do the annual ordering and holding costs compare?
- State the newsvendor optimality condition for continuous demand.
Answers
- \( Q^* = \sqrt{\frac{2AD}{h}} \): \( A \) cost per order, \( D \) annual demand, \( h \) holding cost per unit per year.
- They are equal, each \( \frac{1}{2}\sqrt{2ADh} \).
- \( F(Q^*) = \frac{c_u}{c_u + c_o} \).
Level 2: Part B Style
Practice Q1. Annual demand is 2400 units, ordering cost ₹150 per order and holding cost ₹2 per unit per year. With no shortages, the time between orders is
(a) 1 month (b) 2 months (c) 3 months (d) 4 months
Answer: (c)
\( Q^* = \sqrt{\frac{2 \times 150 \times 2400}{2}} = \sqrt{360000} = 600 \). Orders per year \( \frac{2400}{600} = 4 \), so one every 3 months.
Practice Q2. A firm orders twice the economic order quantity each time. Its total annual ordering-plus-holding cost exceeds the minimum by
(a) 12.5% (b) 25% (c) 50% (d) 100%
Answer: (b)
\( \frac{TC}{TC^*} = \frac{1}{2}\left(2 + \frac{1}{2}\right) = 1.25 \). (Half the EOQ gives the same 25%.)
Practice Q3. Demand 1200 units per year, ordering cost ₹100, holding cost ₹6 per unit per year, backorder cost ₹12 per unit per year. The optimal order quantity is approximately
(a) 163 (b) 200 (c) 245 (d) 283
Answer: (c)
\( Q^* = 200\sqrt{\frac{6 + 12}{12}} = 200\sqrt{1.5} \approx 244.9 \). (a) is the maximum inventory \( S^* \); (b) is the no-shortage EOQ; (d) is \( 200\sqrt 2 \).
Practice Q4. A seasonal item costs ₹20, sells for ₹50, and leftover units are sold at ₹5. Season demand is uniform on \( [100, 400] \). The optimal order quantity is
(a) 200 (b) 250 (c) 300 (d) 350
Answer: (c)
\( c_u = 30 \), \( c_o = 15 \); critical ratio \( \frac{2}{3} \); \( 100 + 300 \times \frac{2}{3} = 300 \). (Ignoring the salvage value, \( c_o = 20 \), would give ratio 0.6 and 280, not an option.)
Level 3: Part C Style
Practice Q5. A product has demand \( D = 6000 \) units per year and is produced at \( P = 15000 \) units per year; setup cost ₹200, holding cost ₹4 per unit per year, no shortages. Take a year as 300 working days. Which statements are true?
(a) The optimal lot size is 1000 units. (b) The maximum inventory is 600 units. (c) Each production run lasts 20 working days. (d) The minimum annual cost is higher than it would be with instantaneous replenishment at the same \( A \) and \( h \).
Answer: (a), (b) and (c)
\( 1 - \frac{D}{P} = 0.6 \).
- (a) True. \( Q^* = \sqrt{\frac{2 \times 200 \times 6000}{4 \times 0.6}} = \sqrt{1{,}000{,}000} = 1000 \).
- (b) True. \( 1000 \times 0.6 = 600 \).
- (c) True. \( \frac{Q}{P} = \frac{1000}{15000} = \frac{1}{15} \) year \( = 20 \) working days (cycle \( \frac{1000}{6000} \) year = 50 days).
- (d) False. \( TC^* = \sqrt{2 \times 200 \times 6000 \times 4 \times 0.6} = 2400 \), lower than \( \sqrt{2 \times 200 \times 6000 \times 4} \approx 3098 \) for instantaneous replenishment (whose EOQ is about 775).
Practice Q6. Demand is 1200 units per year, ordering cost ₹100 and holding cost 20% of the unit price per year. The price is ₹10 per unit for orders below 500 and ₹9.60 for orders of 500 or more (on all units). Which statements are true?
(a) The EOQ at the price ₹10 is about 346 units. (b) The EOQ at the price ₹9.60 is about 354 units, so this price is not available at its own EOQ. (c) The optimal policy is to order 500 units. (d) The minimum total annual cost, including purchase, is ₹12,240.
Answer: (a), (b), (c) and (d)
- (a) True. \( h = 2 \): \( \sqrt{\frac{240000}{2}} = \sqrt{120000} \approx 346.4 \).
- (b) True. \( h = 1.92 \): \( \sqrt{\frac{240000}{1.92}} = \sqrt{125000} \approx 353.6 < 500 \).
- (c) True. Candidates: \( Q = 346.4 \) at ₹10 with total cost \( 12000 + 692.8 = 12692.8 \); \( Q = 500 \) at ₹9.60 with total cost \( 11520 + 240 + 480 = 12240 \). The second is cheaper.
- (d) True. As computed, ₹12,240.
All four options are correct: in Part C this is allowed, and you must mark all four to score.
Practice Q7. For the basic EOQ model, which statements are true?
(a) If annual demand quadruples (other parameters fixed), the EOQ doubles. (b) If the ordering cost and the holding cost both double, the EOQ is unchanged and the minimum annual cost doubles. (c) Ordering \( 1.25 \) times the EOQ raises annual cost by 25%. (d) If shortages are allowed at a finite backorder cost, the optimal order quantity is smaller than the EOQ.
Answer: (a) and (b)
- (a) True. \( Q^* \propto \sqrt{D} \).
- (b) True. \( \frac{A}{h} \) unchanged, so \( Q^* \) is unchanged; \( TC^* = \sqrt{2ADh} \) scales by \( \sqrt{2 \times 2} = 2 \).
- (c) False. \( \frac{1}{2}\left(1.25 + \frac{1}{1.25}\right) = \frac{1}{2}(1.25 + 0.8) = 1.025 \): only 2.5% more.
- (d) False. The factor \( \sqrt{\frac{h + s}{s}} > 1 \) makes it larger.
Quick Revision Table
| Model | Optimal lot | Minimum cost |
|---|---|---|
| EOQ | \( \sqrt{\frac{2AD}{h}} \) | \( \sqrt{2ADh} \) |
| With backorders | \( \sqrt{\frac{2AD}{h}}\sqrt{\frac{h + s}{s}} \); \( S^* = \frac{s}{h + s}Q^* \) | \( \sqrt{2ADh}\sqrt{\frac{s}{h + s}} \) |
| Production (EPQ) | \( \sqrt{\frac{2AD}{h(1 - D/P)}} \); \( I_{\max} = Q(1 - \frac{D}{P}) \) | \( \sqrt{2ADh(1 - \frac{D}{P})} \) |
| Discounts | EOQ per price, move to break point if below range, compare total cost | includes \( Dc \) |
| Newsvendor | \( F(Q^*) = \frac{c_u}{c_u + c_o} \) (discrete: smallest \( Q \) with \( F \ge \)) | |
| Reorder point | \( R = DL \) (inventory position) | |
| Sensitivity | \( \frac{TC}{TC^*} = \frac{1}{2}(\frac{Q}{Q^*} + \frac{Q^*}{Q}) \) |
Key Takeaways
What’s Next
- Reliability and Operations Research: Quick Revision: every formula of the chapter on one page.
- Queuing Models: the other elementary stochastic OR model in the syllabus.
- Linear Programming: Simplex and Duality: constrained versions of these problems (several items sharing storage or budget) become mathematical programs.