Safety Stock Calculator
Work out safety stock and the reorder point from daily demand, lead time and the service level you want, with the common formulas side by side. Every result shows the working, and the methods your inputs allow appear next to each other.
Your numbers
Units sold or used on an average day, over the last few months.
The busiest normal day you plan for, not a one-off spike.
Days from placing an order to having the stock on the shelf.
The slowest delivery you have seen recently.
The chance of not running out before the next delivery arrives (cycle service level). 50 to 99.99.
In units per day. A spreadsheet gives it: =STDEV.S() over your daily sales.
In days, from your past deliveries. Leave empty if the lead time hardly varies.
Safety stock
–
The headline uses the most complete method your inputs allow. Safety stock and reorder point are rounded up to whole units.
All fields use the same time unit. For weeks, enter weekly demand and lead time in weeks.
The calculation runs in your browser and ParrotNotes does not store your numbers. They sit in the page address so you can bookmark or share the result.
The safety stock formulas, with one example
Take an item that sells 40 units on an average day and 60 on a busy one. Orders usually arrive in 10 days, the slowest took 14. From the last few months of sales, the standard deviation of daily demand is 8 units; from past deliveries, the standard deviation of lead time is 2 days. You want a 95% service level, so z = 1.645. Demand over an average lead time is 40 × 10 = 400 units.
Max minus average: 60 × 14 − 400 = 440 units. Demand varies: 1.645 × 8 × √10 = 41.6, so 42 units. Demand and lead time vary: 1.645 × √(10 × 8² + 40² × 2²) = 1.645 × √7,040 = 138.0, so 139 units. The reorder points are 840, 442 and 539.
The spread between 42 and 440 is the point of comparing. The second formula ignores late deliveries, and here they are the bigger risk: 40² × 2² = 6,400 of the 7,040 comes from lead time. The first formula plans for the worst day and the worst delivery at once, which rarely happen together.
Which method to use
Max minus average needs only four numbers you can read off an order history, which is why it is common in small businesses. It has no service level behind it: you cannot say how often it runs out, and one unusual day or delivery sets the whole buffer.
z × σd × √LT is the textbook formula when the supplier is reliable. It assumes each day's demand varies independently and roughly follows a normal distribution, so the spread over the lead time grows with the square root of its length.
z × √(LT × σd² + d² × σLT²) adds lead-time variation. It assumes demand and lead time vary independently of each other and that orders do not overtake one another. If your supplier's lead time moves at all, this is the one to start from.
Service level and z
The service level here is the cycle service level: the chance of not running out between placing an order and receiving it. z is the point on the standard normal distribution below which that share of outcomes falls, computed in the page rather than read off a table: 1.282 for 90%, 1.645 for 95%, 2.054 for 98%, 2.326 for 99% and 3.090 for 99.9%.
A cycle service level is not a fill rate. At 95% you expect to run short in about one replenishment cycle in twenty, but the units short in that cycle are usually a small share of everything sold, so the share of demand filled from stock is higher than 95%. If a customer contract states a fill rate, this formula with the same percentage will hold more stock than the contract needs.
Getting the inputs right
Use daily sales or usage for the standard deviation, not monthly totals, and leave out days the item was out of stock: a zero on those days is missing demand, not low demand. In a spreadsheet, =STDEV.S() over a column of daily figures gives σd. Do the same over the lead times of your last deliveries for σLT.
If you forecast demand, the forecast error is a better input than the standard deviation of sales, because the buffer has to cover what you did not predict. MIT's supply chain course notes treat the standard deviation of demand as a reasonable stand-in when the error is not tracked.
Keep one time unit throughout. If you order weekly and think in weeks, enter weekly demand, its weekly standard deviation and the lead time in weeks.
What this calculator does not cover
It assumes demand is steady over the season you plan for. Seasonal or trending items need the figures from the period ahead, not last year's average. Lumpy demand, such as a few large orders a month with nothing in between, is not well described by a normal distribution, and the result will be off.
It is for continuous review: you check stock all the time and order when it hits the reorder point. If you only order on a fixed day each week or month, the stock has to cover the review period as well as the lead time, so use lead time + review period in the lead time fields.
Safety stock, EOQ and lead time
This page is about the buffer: how much stock to hold against uncertainty and which method gives which figure. The EOQ calculator works out how much to order each time, and includes a simple safety stock for steady lead times. The lead time calculator counts the business days from order to delivery, which is where the average lead time here comes from.
Safety stock sits in your inventory all year, so it lowers inventory turnover. The inventory turnover calculator shows what the whole stock turns, buffer included.
For distributor and manufacturers' rep conversations
When a distributor's buyer says they ran out of your line, the question is usually which number moved: demand or your lead time. Bring their sales history and your delivery record, run both formulas and show what a two-day improvement in lead-time spread would save them in stock. That is a conversation about their cash, not your price.
Sources, each read on 2026-11-08. Max minus average is a rule of thumb from inventory practice with no single source. Results are arithmetic on your own numbers:
- Stanford MS&E 262 lecture notes (Hausman, Kleinknecht and Wilkinson): demand over a random lead time has variance μ²s² + σ²τ, if orders do not cross
- MITx CTL.SC1x, Probabilistic inventory models: reorder point = expected lead-time demand + k × σ; cycle service level = P[no stockout in a cycle]
- NIST/SEMATECH e-Handbook, Cumulative distribution function of the standard normal: the table the z values are checked against
Related: the EOQ calculator for the order size, the lead time calculator for the days from order to delivery, and the inventory turnover calculator for what the whole stock turns.
More for distributor and manufacturing sales: Vendor managed inventory, Distributor sales, Industrial distributor.
The number starts the stocking talk
The buffer shows what late deliveries cost; the call with the buyer or the supplier decides what changes. Record that call in ParrotNotes and you get a transcript and a summary, so what was agreed is written down.
Get Started FreeRelated tools
- EOQ CalculatorWork out the economic order quantity from yearly demand, order cost and holding cost, with orders per year, the reorder point, safety stock and total cost.Open the tool
- Lead Time CalculatorCount business days from order to delivery, find the order-by date for a needed date, and add up supplier, production and shipping lead times.Open the tool
- Inventory Turnover CalculatorWork out inventory turnover from cost of goods sold and average inventory, with days on hand and a check against your own target.Open the tool
- Days Sales in Inventory CalculatorWork out days sales in inventory (DSI) from average inventory and cost of goods sold, with inventory turnover for the same period.Open the tool