Gains:
- Ability to calculate EOQ, safety stock and reorder point concepts
- Ability to use AI as an aid in ABC/XYZ analysis and inventory policy design
- Ability to verify AI suggested stock parameters with formula and actual consumption data
Inventory is the frozen money of the business. Too much stifles cash flow, creates warehouse costs and wastage; Less than that means loss of sales, line stoppages and missing customers. The task of the industrial engineer is to establish this balance not with "feeling" but with formula and data. Artificial intelligence accelerates accounting, classification and policy design in this area; However, stock formulas are very sensitive to unit and assumption errors, so the AI output must be verified by recalculation. In this unit, we will cover EOQ, safety stock, reorder point and ABC/XYZ analysis in an AI-assisted manner.
Basic Concepts and Formulas
Inventory management has three core questions: How much should I order?, When should I order?, How much buffer should I keep?
Economic Order Quantity (EOQ): The quantity that balances the cost of ordering and the cost of holding.
EOQ = square root( (2 × D × S) / H )D = annual demand (unit/year)S = cost per order (TL/order)H = annual holding cost per unit (TL/item/year)
Reorder Point (ROP): When the stock drops to this level, a new order is placed.
ROP = d × L + SSd = average daily (or periodic) demandL = lead time (in the same unit of time)SS = safety stock
Safety Stock (SS): Buffer against uncertainty in demand and lead time.
SS = Z × σ_d × square root(L)Z = confidence coefficient corresponding to the service level (e.g. ≈ 1.65 for 95%)σ_d = standard deviation of demand (per period)L = lead time (number of periods)
Caution: The most common error in these formulas is unit inconsistency. If you give the demand daily and the supply time weekly, the result will be meaningless. Convert all entries to the same time unit before starting each account. AI often skips this transformation.
Sample Calculation: Step by Step
For a spare part, let D = 3,600 pieces/year, S = 120 TL/order, H = 8 TL/piece/year.
EOQ = square root( (2 × 3600 × 120) / 8 ) = square root( 864000 / 8 ) = square root( 108000 ) ≈ 328.6 → about 329 unitsNumber of orders per year = 3600 / 329 ≈ 10.9 ≈ 11 orders
Now ROP: daily demand d = 3,600 / 300 working days = 12 units/day, lead time L = 10 days, demand standard deviation σ_d = 4 units/day, target service level 95% (Z ≈ 1.65):
SS = 1.65 × 4 × square root(10) = 1.65 × 4 × 3.162 ≈ 20.9 → 21 unitsROP = 12 × 10 + 21 = 141 units
In other words, when the stock drops to 141 units, a new order of 329 units is placed. When you outsource this type of calculation to AI, ask it to display each row open like this; An output that says "result is 141" cannot be verified.
Focus with ABC and XYZ Analysis
Managing every item with the same care is a waste. ABC analysis classifies items by value (annual usage × unit cost): A items (usually ~20% of items, ~80% of value) are tightly managed. XYZ analysis classifies according to demand variability: X is regular, Z is irregular demand.
class
Meaning
Politics
A-X
High value, regular demand
Tight tracking, low safety stock, frequent ordering
A-Z
High value, erratic demand
Close monitoring, careful buffering, flexible supply
C-X
Low value, regular demand
Simple policy, automatic ordering
C-Z
Low value, irregular
Take as needed, low priority
Role: You are an industrial engineer specializing in inventory management.Task: Insert the following list of items into the matrix ABC (annual usage value) and produce
Weak Prompt / Strong Prompt
Weak prompt:
What should be the safety stock for these products?
Without service level, demand variability and lead time, any answer is a guess.
Powerful prompt:
Calculate safety stock and ROP for the following item. Service level 98%(specify Z value), average daily demand 30, daily demand std deviation9, lead time 7 days. Write the formula, show each step, check the units. Then compare how SS changes if I reduce the service level to 95%.
The second prompt contains both the calculation and a "sensitivity" question; This allows you to see how much the decision depends on which assumption.
Mini Case: The Cost of Service Level
An automotive supply industry company aims for a 99.9% service level to never leave a critical gasket missing. Baran, an industrial engineer, has the AI calculate safety stock for current and alternative service levels. The formula shows that going from 95% to 99.9% moves Z from 1.65 to approximately 3.09; so the safety stock almost doubles. When Baran multiplies this by the item's holding cost, he sees that that last percent is very expensive. It reduces the total inventory cost by selecting 95% for non-critical items and 99% for critical items that stop the line. AI set up account; It was Baran's engineering judgment that decided what level of service was worth the work.
Hint: Safety stock grows exponentially, rather than linearly, with service level. The last percentiles (99% → 99.9%) are very expensive. Question whether each item really needs that high level.
Common Mistakes
- Unit mixing: Entering demand daily and supply time weekly and finding meaningless SS/ROP.
- Single service level: Applying the same service level to all items and managing items A and C equally.
- Not validating the AI calculation: Accepting EOQ/SS results without seeing the square root and multiplication steps.
- Constant demand assumption: Ignoring variability (σ) and keeping the safety stock at zero.
- Ignoring dead stock: Continuing to place orders with the old parameters for items with decreased demand.
In summary
- Stock answers three questions: how much (EOQ), when (ROP), how much buffer (SS). Each has a clear formula.
- Unit consistency is the most critical and often overlooked point of inventory accounting; Be sure to verify the AI output with a unit check.
- The ABC/XYZ matrix directs effort to the right items; It is a waste to manage every item with the same care.
- As the level of service increases, safety stock quickly becomes expensive; Query the last percentiles.
- AI establishes and speeds up the account; It is human judgment that decides which policy is worth the work.
Application task
Prepare a stock list of at least 8-10 items (code, annual demand, unit cost, demand std deviation, lead time). First have the AI construct the ABC/XYZ matrix and clearly display the coefficient of variation. Then calculate the EOQ, safety stock and ROP values of an A-item; See every step and unit. Critical verification: Redo the EOQ and SS calculation in person with Excel or a calculator and compare with the AI. Finally, calculate the service level for the same item in the 95% and 99% scenarios and interpret the safety stock difference and its annual cost impact.