Two separate questions
Inventory decisions look like one problem and are two. When do I place the order, and how much do I put on it. The first is answered by the reorder point, the second by the order quantity, and they are computed from different things.
The reorder point is the stock that will be sold while you wait for the delivery, plus the buffer. Forty units a day and a seven-day lead time means 280 units go out before anything arrives; five days of safety stock is another 200. Reorder at 480. Wait until 300 and the shelf empties before the truck comes, regardless of how large the order was.
How many days of safety stock
Safety stock covers two different failures: selling faster than average, and receiving later than promised. How thick it needs to be depends on how erratic each of those is for the item in front of you. Something that sells at a steady rate from a supplier who has never been late needs very little. Something that sells in bursts, arriving by sea, needs a lot.
The price of the buffer is on the other side of the balance sheet: cash tied up, space occupied, and — for anything dated or seasonal — the risk of writing it off. That trade is why the answer differs by item rather than by business. A high-margin item where a stockout means a lost customer justifies a deep buffer. A bulky low-margin one does not. Whether stock overall is moving or sitting is a different measurement, on the inventory turnover calculator.
Order quantity, case size and what rounding does
The quantity here fills you back to a target level: lead time plus safety stock plus however many days each order should add. Set that last field to thirty and you order roughly monthly; set it to fourteen and you order fortnightly. Small frequent orders tie up less cash and cost more in freight and handling. Large infrequent ones do the reverse.
Cases blunt the result. Needing 1,160 units with 24 to a case is 48.3 cases, so you buy 49 and take 1,176. The overshoot is trivial here and enormous on a slow item — a case of 24 against sales of two a week is three months of stock in one purchase, and no reorder point can fix that. When the case size is a large multiple of weekly sales, the case size is the ordering policy and the calculation is just describing it.
The demand multiplier exists for the weeks when the average is known to be wrong. Going into a busy season, 1.4 raises the reorder point and the target together, so the order placed before the season is sized for the season. Ordering at the normal rate on the way out of one is how a stockroom fills up.
EOQ, and the assumptions underneath it
The economic order quantity is the order size where the annual cost of placing orders and the annual cost of holding stock add up to the least. The formula is the square root of 2DS/H: D is annual demand in units, S is the cost of placing one order, H is the cost of holding one unit for a year.
Work one through. Demand of 10,000 units a year, $40 to place an order, $2 a unit a year to hold it. Two times 10,000 times 40 is 800,000; divided by 2 is 400,000; the square root is 632.5, so about 632 units an order. That is 15.8 orders a year, costing 15.8 × $40 = $632 in ordering, and an average of 316 units on hand costing 316 × $2 = $632 to hold. The two halves come out equal, which they always do at the optimum — that is what the formula is finding.
Now the assumptions: demand is constant across the whole year, lead time never varies, the unit price is the same at any quantity, and nothing limits how much you can store or how long you can keep it. Those rarely hold, and they rarely fail one at a time. A seasonal product breaks the first, a supplier with pallet pricing breaks the third, and anything with a date code breaks the fourth. Use EOQ to notice that you are ordering ten times too often or ten times too much. Do not use it as the order quantity.
Volume discounts and the arithmetic that usually loses
The most common way a stockroom gets overfull is a price break. Ten percent off for triple the quantity is a real saving on paper, and the costs on the other side — months of holding, the cash unavailable for something else, the units that go out of date or out of fashion — are diffuse and easy to leave out. Put both sides in and the discount often loses, particularly on anything where demand might change. The comparison is laid out properly on the bulk buy optimizer, and what the item earns per unit sits on the margin calculator.
Where this model stops describing your business
It assumes an item with a history. A new product has no daily rate to start from, and borrowing one from a similar item is a guess wearing a number's clothes. The usual approach is a deliberately small first order and a real rate computed after two or three weeks of sales, accepting an early stockout in exchange for not owning a pallet of something nobody wanted.
It also assumes items are independent. If a supplier has a minimum order value or free freight above a threshold, the right decision is about the whole purchase order rather than one line, and the reorder points here become inputs to a bigger question about what else to pull forward. And it assumes the sales history is demand — a period that included a stockout recorded fewer sales than there was demand, so feeding that period back in makes you order less and stock out again. When the numbers keep coming out low, check whether you are measuring what you sold or what people wanted.
Questions people ask
What period should the daily sales figure come from?
Four to eight weeks of sales divided by the days in that period is a reasonable default. Shorter than that and one promotion or one large customer distorts it; longer and it lags a real change in demand. If your week has a strong shape — busy Saturdays, dead Mondays — total whole weeks and divide by seven rather than picking days, so you never start from a partial week. One caution: if the period contained a stockout, the sales recorded are lower than the demand that existed, and using them will size your next order too small.
My lead time varies a lot. What do I enter?
Not the average. The reorder point has to survive the slow deliveries, so use something near the bad end of what you actually see. If it is usually seven days and occasionally fourteen, either enter ten and keep normal safety stock, or enter seven and increase the safety days — both raise the reorder point, which is the thing that protects you. The scenario table on the page includes a case with the lead time seven days longer, so you can see how far the reorder point moves before choosing.
Should stock already on order count as on hand?
Not in the on-hand field, but it does have to be counted somewhere or you will order the same thing twice. The standard practice is to compare the reorder point against on hand plus on order — inventory position rather than physical stock. The simplest way to use this page is to add the quantity already ordered into the on-hand figure and keep a separate note of when it lands, since the calculator has no field for a delivery date and will otherwise assume everything is sellable today.
Why is the EOQ so different from the recommended order quantity?
They answer different questions. The recommended quantity fills you back to a target stock level given what is on the shelf right now, so it depends on your current position and the ordering rhythm you chose. EOQ ignores your current position entirely and asks what order size makes ordering costs and holding costs add up to the least over a year. Similar magnitudes mean your rhythm is roughly sensible. An EOQ far larger usually means the cost per order is set high or the holding cost low, and if the item has a shelf life you cannot follow it anyway.
How do I set the annual holding cost per unit?
Add up what a unit actually costs you to keep for a year: the storage it occupies, the return you could have had on the cash while it sat there, insurance, and the realistic loss to damage, theft and obsolescence. Businesses often express this as a percentage of unit cost, but the percentage varies enormously — a pallet of dated food and a shelf of small metal parts are not remotely the same problem — so build it from your own numbers rather than adopting one. It only affects the EOQ line here, so an approximate figure is fine as long as it is not zero.