Inventory Reorder Point Calculator
Daily sales multiplied by lead time is the quantity you expect to sell while the container is at sea. It is the level at which you run out when demand behaves exactly as forecast — which it does about half the time. The reorder point sits above it by the safety stock, and the safety stock is the only part of your inventory that exists purely to absorb being wrong.
| Building the reorder point | Value | Notes |
|---|---|---|
| Expected sales during lead time | — | — |
| Standard deviation of demand over lead time | — | — |
| z for your service level | — | — |
| Safety stock | — | — |
| Reorder point | — | — |
| Days of cover at reorder point | — | — |
| What the buffer costs you each year | Amount |
|---|---|
| Average inventory (safety + half an order) | — |
| Cash sitting in average inventory | — |
| Holding cost on average inventory | — |
| Replenishment cycles per year | — |
| Probability of running short in a cycle | — |
| Expected stockout events per year | — |
| Annual stockout cost | — |
| Holding + stockout cost per year | — |
| Cost of moving to the cheapest service level | — |
| Service level | z | Safety stock | Holding / yr | Stockouts / yr | Stockout cost | Total / yr |
|---|
| Lead time | Demand in transit | Safety stock | Reorder point | Holding / yr |
|---|
| Order quantity | Cycles / yr | Cycle stock | Holding / yr | Cheapest service level |
|---|
The reorder point is not where you order. It is where you would run out.
At the default figures — 30 units a day, a 20-day lead time — you will sell 600 units between placing the order and the container arriving. Set the reorder point at 600 and the stock hits zero on the same day the delivery lands, on average. Since demand is above average roughly half the time, you spend about half of those cycles waiting for a shipment with nothing to sell.
Demand over the lead time is not a fixed 600. At a standard deviation of 9 units a day, the spread over 20 days is 9 × √20 ≈ 40.2 units — not 180, because daily fluctuations partly cancel out. Cover 95% of that spread and you need 66 units of safety stock, which puts the reorder point at 666 units.
| Component | Units | What it is for |
|---|---|---|
| Demand during lead time | 600 | Sells while the order is in transit |
| Safety stock (95%) | 66 | Absorbs demand running above average |
| Reorder point | 666 | Place the order here, not at 600 |
The 66 units are not free: at $18 a unit that is about $1,190 of stock standing still. Whether that is worth paying depends entirely on how often your supplier is late and how much a stockout costs you — which is what the rest of this page is about.
The last point of service level is where the money disappears
Service level is usually chosen out of habit. 95% sounds careful, 99% sounds professional, 99.9% sounds like Amazon. Here is what each one actually costs at the default figures, with a $175 stockout cost and 18.25 replenishment cycles a year:
| Service level | Safety stock | Holding / yr | Stockouts / yr | Stockout cost | Total / yr |
|---|---|---|---|---|---|
| 80% | 34 | $1,502 | 3.65 | $639 | $2,141 |
| 85% | 42 | $1,538 | 2.74 | $479 | $2,017 |
| 90% | 52 | $1,582 | 1.83 | $319 | $1,901 |
| 95% | 66 | $1,648 | 0.91 | $160 | $1,808 |
| 97.6% | 80 | $1,709 | 0.44 | $76 | $1,785 |
| 99% | 94 | $1,771 | 0.18 | $32 | $1,803 |
| 99.5% | 104 | $1,817 | 0.09 | $16 | $1,833 |
| 99.9% | 124 | $1,910 | 0.02 | $3 | $1,913 |
Read it in three steps. Going from 80% to 95% costs $145 more in holding and saves $479 in stockouts, so the move is worth $334 a year. Going from 95% to 99% costs $123 and saves $128 — a wash, and you could argue either way. Going from 99% to 99.9% costs $138 and saves $29. You are buying protection against 0.16 stockout events a year at roughly $685 per event avoided, against your own estimate that an event costs $175.
That is the shape of every service-level decision: the protection gets more expensive faster than the risk gets smaller. The cheapest point here is 97.6%, worth $23 a year over 95% — a difference small enough that it is not worth arguing about, which is itself the finding. Between 90% and 99% you are haggling over about $120 a year. Below 90% you are losing hundreds.
Double the lead time, and the buffer grows by 41%
Safety stock moves with the square root of lead time, because independent daily fluctuations partially cancel. Keep the service level at 95% and stretch the supplier:
| Lead time | Demand in transit | Safety stock | Reorder point | Holding / yr |
|---|---|---|---|---|
| 10 days | 300 | 47 | 347 | $1,561 |
| 20 days | 600 | 66 | 666 | $1,648 |
| 40 days | 1,200 | 94 | 1,294 | $1,771 |
| 60 days | 1,800 | 115 | 1,915 | $1,866 |
| 80 days | 2,400 | 132 | 2,532 | $1,946 |
From 20 days to 80 days, the reorder point goes up 3.8× but the safety stock only doubles. Of the 1,866 extra units you have to fund, 1,800 are goods in transit and 66 are buffer.
This is the part that gets misread. A longer lead time is not primarily a service-level problem — it is a working-capital problem. At 80 days you are paying for 2,400 units to be somewhere between the factory and your warehouse, and if payment is on shipment that is about $43,000 tied up in goods you cannot sell yet, against $10,800 at 20 days. Cutting lead time from 80 to 40 days frees roughly $21,600 in cash and needs no negotiation on price.
Lead time variability hits harder than average lead time, which is why the calculator takes it as a separate input. A supplier who averages 20 days but swings between 14 and 30 is worse than one who reliably takes 24: the spread enters the formula as √(LT × σ_daily² + daily² × σ_leadtime²), and the second term scales with the square of daily sales, so it dominates at volume.
Variability is a permanent bill, not a one-off risk
Hold the service level at 95% and change only how erratic demand is:
| Demand variability | Safety stock | Holding / yr | Stockouts / yr |
|---|---|---|---|
| σ = 4.5 (CV 15%) | 33 | $1,499 | 0.91 |
| σ = 9 (CV 30%) | 66 | $1,648 | 0.91 |
| σ = 18 (CV 60%) | 132 | $1,946 | 0.91 |
| σ = 27 (CV 90%) | 199 | $2,244 | 0.91 |
Note the last column: it never moves, because service level is a decision, not an outcome. What variability changes is the price of the decision. A store with 90% demand variability pays $745 a year more than one at 15% for exactly the same protection.
This matters because the two levers are not equally priced. Buying more buffer is a recurring annual cost forever. Reducing variability — a shorter review period, better forecasting, holding a fast-mover separately instead of averaging it with a slow one, promoting in ways you can predict — is a one-off process change. Where the buffer is expensive, the second lever earns its keep.
Big order quantities should not run at high service levels
One thing this model gets right that a formula-only page misses: the cheapest service level depends on how often you replenish, not just on cost and risk.
| Order quantity | Cycles / yr | Cycle stock | Holding / yr at 95% | Cheapest service level |
|---|---|---|---|---|
| 150 | 73.0 | 75 | $635 | 99.5% |
| 300 | 36.5 | 150 | $973 | 98.9% |
| 600 | 18.3 | 300 | $1,648 | 97.6% |
| 1,200 | 9.1 | 600 | $2,998 | 94.4% |
A stockout can only begin when stock is being consumed from an open cycle, and each cycle is one exposure. Ordering 1,200 units instead of 150 cuts the number of cycles from 73 to about 9, so the risk is tested eight times less often and the extra protection buys correspondingly less. Freight-driven bulk buying and a 99% service target are pulling in opposite directions, and the bill is paid in holding cost.
The reverse is also true and more useful: splitting a 600-unit order into two 300-unit orders halves cycle stock, saves $675 a year in holding, and raises the cheapest service level to 98.9% at the same time. The reason not to do it is ordering cost and freight minimums, which this model does not include — check those before acting on it.
What a stockout actually costs you
The cost per stockout event is the input most people get wrong, and it is the one that moves the answer most. It is not the lost gross profit on the units you could not ship.
| Component | Example | Amount |
|---|---|---|
| Gross profit lost on the short units | 3 units × $21.99 margin | $66 |
| Customers who do not come back | 35% of them × $90 lifetime profit | $32 |
| Compensation or discount code | Customer service gesture | $20 |
| Premium to expedite a replacement | Air freight difference per event | $60 |
| Cost per event | — | $178 |
That is where the $175 default comes from. Change it and the cheapest service level moves with it, because the whole calculation is a comparison between two bills:
| If a stockout costs | Cheapest service level | Safety stock | Why |
|---|---|---|---|
| $50 | 88.1% | 48 | Only the lost unit margin is counted |
| $100 | 95.3% | 67 | Roughly margin plus a replacement order |
| $175 | 97.6% | 80 | Margin plus a lost customer |
| $300 | 98.7% | 90 | A repeat buyer with real lifetime value |
| $900 | 99.6% | 108 | A wholesale account you cannot replace |
If you price a stockout at $50, the model tells you to run at 88% service — which is right, if the only thing you lose is the margin on three units. If the same stockout costs you a wholesale account, the answer is 99.6%. Both are correct answers to different questions, and the common failure is using the $50 answer while carrying the $900 risk.
Where this model stops working
Two parameters and a normal distribution carry a lot of assumptions. These are the cases where the output is not worth trusting:
- Seasonal demand. A single average and a single standard deviation describe the whole year, so a product that sells 10 a day in January and 90 in December gets a buffer that is wrong in both months. Split the year into seasons and calculate each separately, or use a forecast-driven reorder point.
- Promotions. A campaign creates a demand spike that is not noise around the average, it is a different regime. Either pre-build stock for the event or exclude promotional weeks from the standard deviation and forecast the event separately — including them inflates the buffer for the other 50 weeks.
- MOQ constraints. If the supplier will only ship 500 units, the calculated order quantity cannot be applied. The formula still gives the right reorder point, but your cycle stock is fixed and the cheapest service level shifts accordingly — see the order-quantity table above.
- Correlated products. If eight SKUs are variations of one product and customers buy whichever size is available, their demand moves together. Treating them as eight independent items understates the buffer each one needs.
- Service level versus fill rate. This calculator targets cycle service level: the probability that you do not run short at all in a cycle. A store can run at 95% cycle service and still fill 99% of units ordered, because the shortfall is usually a few units at the end of a cycle. Be explicit about which one you are promising customers.
- In-transit inventory. Goods on the water are not in the holding cost here, which is the standard treatment but understates the true cost of a long lead time. Add
daily × lead time × unit costto the working capital figure yourself if payment is on shipment.
Frequently asked questions
My average daily sales is a guess. How much does that matter?
A lot, and asymmetrically. If the average is too low you under-order and stock out repeatedly; if it is too high you carry a dead buffer. The standard deviation matters even more for the safety stock, because the safety stock is proportional to it. Use at least 8–12 weeks of daily unit sales excluding promotional spikes, and if the number has been drifting, use the recent weeks rather than a two-year average that blends a different business.
Why is the safety stock smaller than one day of sales in my case?
It can be, and that is not a bug. Safety stock protects against the spread of demand around the average, not against the average itself. A product with 30 units a day and only 3 units of daily variation over a 10-day lead time needs very little buffer — the pipeline covers the average, the buffer only covers being wrong. If the answer looks too small, check whether the standard deviation you entered is realistic; a common mistake is entering the standard deviation of monthly sales instead of daily.
What if the lead time varies a lot?
Enter it. The model combines both sources as √(lead time × σ_daily² + daily² × σ_leadtime²), which is the standard treatment for independent variation. The second term grows with the square of your daily sales, so a volatile lead time hurts a high-volume product far more than a slow one. A supplier whose transit time swings by a week is often worse than one who is reliably three days slower.
Should I use 95% or the cheapest service level the calculator shows?
Use the cheapest one as the reference point and treat the difference as the price of a preference. At the defaults, 97.6% is cheapest and 95% costs $23 a year more — a rounding error, so 95% is fine. Where the gap is large, usually because the stockout cost is much higher than the defaults, the calculation is telling you something worth acting on. What is almost never right is picking 99.9% because it sounds safe; it is the most expensive row in the table.
How do I apply this across a whole catalogue?
Run it per SKU, never on the aggregate. Safety stock is driven by variability and stockout cost, both of which differ wildly inside one catalogue: a cheap slow mover and an expensive fast mover with the same average sales need completely different buffers. A practical shortcut is to classify by revenue and variability, calculate the buffer for the few items that carry most of the money, and let the tail sit at a lower service level — the same arithmetic that makes 99.9% a bad idea for individual SKUs applies even more strongly across a long tail.
Does this account for the cash cost of holding the buffer, or just the warehouse cost?
Both, combined into one annual holding rate. The 25% default is a common convention covering storage, insurance, shrinkage, obsolescence and the cost of capital — for a small seller with borrowed money or a credit line, the true rate is often higher. If yours is 40%, the cheapest service level will fall, because every unit of buffer costs more. Adjust the field rather than adding costs elsewhere.
Last reviewed: October 2026. Assumes stationary demand and independent daily variation; excludes ordering cost, freight minimums, in-transit working capital, seasonality and promotions, which are covered under the limitations above. Figures are arithmetic on your inputs, not a forecast of your demand.