Real-World Mathematics: Inventory, Demand, Reorder Points and Stock Flow

Application of Mathematics in Real-World Usage · Worked application 12 · BTT Mathematics Hub

A storeroom contains 120 units. Is that a lot? The question cannot be answered without knowing demand, replenishment time, variability and what happens if stock runs out. One hundred and twenty units may be months of supply for a slow item or less than one day for a fast one.

Inventory turns arithmetic into a time-dependent system. Stock is a quantity held at an instant. Demand is a flow that reduces it. Replenishment arrives after a delay. A reorder point is a trigger, not an order quantity. Safety stock is a buffer, not free capacity. These distinctions make inventory a useful application of rates, inequalities, sequences, probability and optimisation.

Every product, demand pattern, cost and lead time below is fictional and designed for Mathematics teaching. This is not business, procurement, financial or supply-chain advice. Real inventory policies depend on actual data, service requirements, supplier constraints, shelf life, capacity, contracts and risk.

The inventory balance is a conservation equation

Over a stated interval, ending stock = beginning stock + receipts − issues − losses, provided all terms use the same unit and boundary. Suppose a store starts with 200 units, receives 150, issues 240 and records 10 damaged units. Ending physical stock is 200 + 150 − 240 − 10 = 100 units.

The equation is an identity about movement across a boundary. It does not explain why 10 units were damaged or whether every issue was recorded correctly. If a physical count says 92 rather than 100, the 8-unit discrepancy is evidence to investigate, not proof of one particular cause.

Always distinguish physical stock, available stock, reserved stock and stock on order when the system uses those categories. Adding “on order” units to physical stock may be useful for a projected position, but it does not make those units available today.

Demand rate converts stock into time cover

Suppose average demand is 12 units per day and 120 units are physically available with no other commitments. At that constant rate, simple days of cover = 120/12 = 10 days.

If demand rises to 20 per day, the same stock covers only six days. The stock quantity did not change; the denominator did. This mirrors travel-time and fuel-use calculations: a stock divided by a flow rate produces time.

Average demand should not be mistaken for guaranteed demand. Ten days of cover at an average rate does not guarantee that stock lasts ten calendar days if daily demand varies.

Lead time is the delay between ordering and usable receipt

Suppose an item averages 15 units per day and replenishment lead time is four days. Expected demand during lead time is 15×4 = 60 units under the constant-rate model.

If stock is reordered only when it reaches 30 units, a four-day lead time would create an expected shortage of 30 units before replenishment arrives, assuming no stock already on order. The trigger occurs too late for the stated model.

Lead time must match the demand time unit. A weekly demand rate multiplied by a lead time in days without conversion is structurally wrong even if the calculator returns a plausible number.

A simple reorder point covers lead-time demand plus a buffer

IBM describes a reorder point as the balance level at which replenishment should be triggered so that stock does not fall below safety stock during lead time. In a deterministic teaching model, use:

Reorder point = expected demand during lead time + safety stock.

If average demand is 15 units/day, lead time 4 days and safety stock 20 units, reorder point is 15×4 + 20 = 80 units.

The calculation does not prove that 20 units is the correct safety stock. It only shows what reorder point follows from that chosen buffer and average-demand model.

Reorder point and order quantity are different controls

A reorder point answers “when should a replenishment order be triggered?” Order quantity answers “how much should be ordered?” Confusing them can produce strange policies.

Suppose reorder point is 80 units and the chosen order quantity is 200 units. Reaching 80 triggers a 200-unit order. It does not mean ordering 80 units.

IBM’s inventory documentation likewise separates reorder point, safety stock and economic order quantity. This separation is mathematically useful even when a business uses a different replenishment method.

Inventory position can include what is already on order

Suppose 70 units are on hand, 40 are already on order and 10 are reserved for known demand. One simple inventory-position convention is on hand + on order − reserved = 70 + 40 − 10 = 100 units.

If the reorder trigger is based on inventory position at 80, no new order is triggered under this convention. If the system instead looked only at on-hand stock, it would trigger one.

The policy variable must therefore be named. “Stock is 70” is incomplete if the decision rule actually uses an inventory position containing open orders and reservations.

A stock ledger reveals timing that totals hide

Start with 100 units. Demand over four days is 20, 35, 25 and 30 units. A replenishment of 60 arrives at the start of day 4. Assume no lost or backordered demand before the arrival event.

Day	Start stock	Receipt	Demand	End stock
1	100	0	20	80
2	80	0	35	45
3	45	0	25	20
4	20	60	30	50

Total starting stock plus receipt is 160. Total demand is 110. Ending stock 50 balances exactly. But the day-by-day ledger tells us something the total cannot: stock approached 20 before the receipt.

If the 60-unit receipt arrived at the end of day 4 instead, only 20 units would be available for 30 units of demand that day. The same total receipts and demand would produce a temporary shortage. Timing changes service even when the end-of-period arithmetic looks similar.

Stockouts require a policy for unmet demand

Suppose 12 units are available and demand is 20. The physical inventory cannot become −8 units. We need a rule for the unmet 8.

In a lost-sales model, 12 units are delivered and 8 units of demand disappear from future obligations. Ending stock is zero and lost sales are 8. In a backorder model, ending physical stock is zero and backorders become 8, to be satisfied later if possible.

The same demand and on-hand stock create different future states under different policies. Negative “inventory” may be used as a bookkeeping representation of backorders, but it should not be confused with negative physical stock.

Safety stock is protection against uncertainty, not a universal constant

IBM defines safety stock as extra inventory held to reduce the risk of stockouts caused by demand, supply or manufacturing variability. IBM also explicitly notes that there is no universal optimal safety-stock formula; the appropriate method depends on the use case and variables.

This is an important mathematical boundary. A school formula such as “two days of average demand” can be used as a policy assumption, but it should not be presented as a generally optimal business rule.

If average demand is 30/day and a teaching policy chooses two days of buffer, safety stock is 60 units. The arithmetic is valid under that policy. Whether 60 produces an acceptable service level depends on actual demand and lead-time behaviour.

Variability matters even when the mean is unchanged

Compare two five-day demand sequences, each with mean 20 units/day. Sequence A: 19, 20, 20, 20, 21. Sequence B: 5, 10, 20, 30, 35.

Both total 100 units, but B has much larger variation. A reorder policy built only from the average may experience different stockout risk under the two patterns.

This is why safety-stock models often use measures of variation. The IBM safety-stock overview discusses formulas using demand, lead-time variation and service targets. Our calculations below keep the statistical assumptions explicit rather than claiming one formula fits every situation.

A simple normal-demand teaching model

For a classroom illustration only, assume daily demand is independent with mean 20 units and standard deviation 4 units, and lead time is fixed at 9 days. Under the independence model, lead-time mean demand is 180 units and lead-time standard deviation is 4√9 = 12 units.

If a chosen service factor is z = 1.645, safety stock under the simple normal approximation is 1.645×12 ≈ 19.74 units, often rounded according to the unit rule. Reorder point would then be about 200 units.

This is conditional on independence, normal approximation, fixed lead time and the meaning assigned to the service factor. If those assumptions fail, the number may not have the claimed interpretation.

Demand variability and lead-time variability can interact

If demand and lead time both vary, uncertainty during replenishment is more complicated than multiplying one standard deviation by one average. Different models combine the sources in different ways depending on assumptions.

Rather than hide that complexity, a first applied-Mathematics treatment can compare scenarios. Suppose daily demand ranges from 18 to 26 units and lead time from 3 to 5 days. Lead-time demand can range from 54 to 130 units under the extreme combinations.

This 54–130 range is not a probability interval. It simply shows the consequences of chosen bounds. If a reorder point of 100 is below the high-demand/high-lead-time scenario, then it cannot guarantee avoiding stockout under all stated scenarios without existing pipeline stock.

The economic order quantity model has a specific objective

A classic EOQ model balances ordering cost against holding cost under restrictive assumptions. Let annual demand be D units, fixed cost per order S and annual holding cost H per unit. In the basic model, annual relevant cost is DS/Q + HQ/2, where Q is order quantity.

Differentiating with respect to Q gives −DS/Q² + H/2. Setting the derivative to zero produces Q* = √(2DS/H).

Take fictional values D = 2,400 units/year, S = $30/order and H = $3/unit/year. EOQ is √(2×2400×30/3)=√48,000 ≈ 219.1 units. If only whole units are possible, nearby integers such as 219 or 220 can be compared directly.

OpenStax notes EOQ as one of the inventory-management tools used to balance inventory-related trade-offs. The simple formula is not universally optimal when discounts, variable lead times, stockouts, capacity constraints or other real features matter.

Why the EOQ optimum occurs where two cost components balance

At Q ≈ 219.1, ordering-cost term DS/Q is about $328.63 per year. Average-cycle holding-cost term HQ/2 is also about $328.63. Their equality follows from the first-order condition in the basic EOQ model.

Order much less, such as Q=100, and ordering cost becomes $720 while holding cost is $150, total $870. Order Q=400 and ordering cost becomes $180 while holding cost is $600, total $780. Near EOQ, total is about $657.27.

The total-cost curve is relatively flat near its minimum, which is operationally useful: an exact-looking square-root result does not necessarily mean 219 is dramatically better than 220.

Price discounts can change the optimisation problem

Suppose unit price is fictional $10 below 500 units and $9.80 at 500 units or more. Purchase cost now changes with Q, so the simple EOQ calculated without price tiers is not sufficient to identify the cheapest annual policy.

The correct comparison evaluates feasible candidate quantities under the full cost function, including purchase, ordering and holding assumptions. A lower unit price can justify higher average inventory—or not—depending on demand and holding cost.

This is a recurring optimisation principle: changing the objective function invalidates an optimum derived from the old function.

Cycle stock and safety stock play different roles

In a basic order-quantity model with steady demand, cycle stock rises after a replenishment and falls toward the next order. Safety stock is an additional buffer intended to absorb uncertainty.

If order quantity is 200 and safety stock 50, a simple sawtooth model may range from about 250 just after receipt to about 50 just before the next receipt. Average on-hand inventory is then roughly Q/2 + safety stock = 100 + 50 = 150 units under the idealised pattern.

Increasing order quantity increases average cycle stock. Increasing safety stock increases the floor. These choices affect inventory differently even if both raise the average.

Perishable inventory adds an age dimension

Suppose 100 units have a remaining usable life of five days, while average demand is only 10 per day. Even if demand is perfectly steady, only about 50 units are expected to be used within five days under a simple first-in-first-out assumption.

Days of cover would report ten days, but shelf life is only five. Holding more stock increases availability while also increasing expiry risk.

Inventory optimisation therefore changes when items perish. A buffer that makes sense for durable components may be wasteful for time-sensitive goods. The model must preserve the product’s actual constraints.

ABC classification is a prioritisation representation

Suppose four fictional items have annual usage values of $50,000, $20,000, $4,000 and $1,000. Total is $75,000. The first item contributes 66.67%; the first two together contribute 93.33%.

A business might choose to monitor high-value items more closely. But any A/B/C thresholds are policy choices. The classification compresses continuous values into categories; it does not change the underlying amounts.

Categories are useful when they route attention, but they can hide meaningful differences within each group. Keep the raw quantities available for decisions near a threshold.

Forecast error should be separated from actual demand

Suppose forecast demand over four weeks is 100, 110, 90 and 120 units, while actual demand is 105, 95, 100 and 130. Errors defined as actual minus forecast are +5, −15, +10 and +10.

Mean error is (+5−15+10+10)/4 = +2.5 units, suggesting a small upward bias over this tiny sample. Mean absolute error is (5+15+10+10)/4 = 10 units, which measures typical magnitude without cancellation.

A near-zero mean error can coexist with large individual misses because positive and negative errors cancel. For safety-stock and service decisions, variability in errors may matter more than signed average error.

Service level and fill rate are not the same metric

Imagine ten order cycles, with nine cycles avoiding any stockout. Cycle service level is 90% in this simplified definition. Now suppose the one stockout cycle misses only 2 units out of 1,000 total units demanded across all cycles. Unit fill rate would be 99.8%.

Both can be correct. One asks how often a cycle experiences any stockout; the other asks what fraction of units are immediately filled. A service target must specify its metric.

This is another denominator problem. Percentages with different denominators should not be compared as if they measured the same outcome.

A complete fictional reorder case

Suppose average demand is 25 units/day, fixed lead time 6 days and chosen safety stock 40 units. Reorder point is 25×6 + 40 = 190 units.

Current on-hand stock is 170, on-order stock 60 and reserved demand 20. Inventory position under the teaching convention is 170 + 60 − 20 = 210 units. If the policy triggers at or below 190, no new order is placed yet.

Two days later, assume 55 units have been issued, reservations remain 20, and the 60 on-order units have not arrived. On-hand becomes 115 and inventory position is 115 + 60 − 20 = 155. The trigger is crossed, so a new order would be indicated by the simplified policy.

This does not prove the policy is optimal. It simply applies the defined rule consistently and shows why on-hand stock alone can give a different signal from inventory position.

Common failure modes

• Stock/rate confusion: inventory units are compared directly with units per day without converting to time cover.
• Lead-time mismatch: days and weeks are multiplied without unit conversion.
• ROP/quantity confusion: the reorder trigger is treated as the amount to buy.
• Pipeline blindness: open replenishment orders are ignored when the policy uses inventory position.
• Negative physical stock: unmet demand is not separated into lost sales or backorders.
• Average-only planning: variable demand is represented only by its mean.
• Universal safety-stock claim: one formula is applied regardless of demand and lead-time behaviour.
• EOQ overreach: a basic square-root optimum is used after its assumptions have changed.

Practice: twenty inventory questions

1. Start 180, receive 90, issue 210, lose 5. Find ending stock.
2. Stock is 144 and demand averages 18/day. Find simple days of cover.
3. Demand is 25/day and lead time is 4 days. Find expected lead-time demand.
4. Add safety stock 30 to question 3. Find reorder point.
5. On hand 80, on order 50, reserved 20. Find inventory position using the stated convention.
6. If ROP is 100, should question 5 trigger a new order?
7. Stock is 12 and demand is 20 with lost sales. Find delivered quantity and lost demand.
8. Repeat question 7 under a backorder policy. State physical stock and backorders.
9. Daily demands 18,20,22,20,20. Find mean.
10. Compare the range of 19,20,20,20,21 with 5,10,20,30,35.
11. Demand mean is 16/day, fixed lead time 9 days. Find mean lead-time demand.
12. Daily demand standard deviation is 3, independent over a fixed 4-day lead time. Find lead-time standard deviation.
13. Using z=1.645 with question 12, estimate safety stock.
14. D=1,800 units/year, S=$40/order, H=$4/unit/year. Find basic EOQ.
15. For Q=200 in question 14, calculate annual ordering-cost and holding-cost terms.
16. Order quantity 300 with safety stock 60. Estimate average on-hand inventory in a simple sawtooth model.
17. Forecasts 50,60,70; actuals 55,50,75. Find signed errors and mean error.
18. Find mean absolute error for question 17.
19. Nine of ten cycles have no stockout. Find cycle service level.
20. Why can a reorder policy be mathematically consistent but still unsuitable in reality?

Worked answers

1. 55 units.
2. 8 days.
3. 100 units.
4. 130 units.
5. 110 units.
6. No, under a trigger at or below 100.
7. 12 delivered, 8 lost.
8. Physical stock zero; backorders 8.
9. 20 units/day.
10. Ranges 2 and 30. Same-looking averages can hide different variability.
11. 144 units.
12. 6 units. 3√4.
13. About 9.87 units. Round only according to the chosen stock-unit policy.
14. About 189.74 units. √(2×1800×40/4)=√36,000.
15. $360 ordering and $400 holding per year.
16. 210 units. 300/2 + 60.
17. +5, −10, +5; mean 0.
18. 6.67 units approximately.
19. 90%.
20. Because real suitability also depends on actual demand distributions, lead-time behaviour, service metric, perishability, supplier constraints, costs, capacity and risk.

Sources and connected learning

For inventory terminology and the role of safety stock, reorder points and EOQ, see IBM: What Is Safety Stock?, IBM: Reordering Inventory Items, and OpenStax: Inventory Management. The demand patterns, costs and policies in this guide are original teaching constructions.

Continue with Construction, Area, Volume and Material Estimation for quantity planning; Transport, Speed, Travel Time and Fuel for stock-versus-rate analogies; and Maps, Bearings, Coordinates and Navigation for spatial representation. Return to the BTT Mathematics Hub.