Safety stock by fill rate and lead-time variability
Counts safety stock in two ways: by cycle service level (the probability of not running out) and by fill rate (what part of demand is met from the shelf), with demand and lead-time variability; shows reorder points and the money in stock.
At a glance
- What question does it answer?
- How much safety stock to hold to meet the required share of demand and how that differs from protecting every order cycle.
- What data is needed?
- Average demand and its weekly spread, lead time and its spread.
- How to read the result?
- Example on the default values: the demand spread over the protection period is 90 units; stock by 95% service level 148 units, by 98% fill rate only 78 units; at the service-level stock the fill rate is 99.6%.
Why and how to reason
Safety stock protects against fluctuations in demand and in lead time. But what exactly do you want to protect? The first answer: the probability of reaching the end of the cycle without a stock-out (cycle service level). The second answer: what share of demand to cover from the shelf (fill rate). These are different goals. A shortage on the last day of the cycle of 5 units and a shortage of 500 units equally spoil the first metric but very differently the second. Protection spread: √((lead time + period) × σdemand² + demand² × σlead²) = √(5 × 900 + 14,400 × 0.25) = 90 units. Safety stock for a 95% service level: 1.645 × 90 = 148. For a fill rate of 98% with an order of 480: G(z) = 0.02 × 480 ÷ 90 = 0.107, giving z ≈ 0.87 and stock of 78. Stock by fill rate is almost half as large and in fact covers 98% of demand.
When to use it
Use it for goods with constant demand and a known lead time, when deciding how much to hold in reserve and when you can say what matters more to you: service or money. Do not use it for novelties without history and for goods with intermittent demand (see Croston). Assumptions: demand and lead time are independent and normally distributed, the spread is measured, orders repeat with one period.
How to read the result
Compare the two stocks: 148 by service level and 78 by fill rate. With stock of 148 the fill rate is 99.6%, with stock of 78 the service level is only 81%. Choose what fits your business: if the absence of even one unit annoys a customer, look at the service level; if the share sold matters, at fill rate. Money in stock (94 thousand) and holding for a year (23 thousand) convert the choice into roubles. The replenishment level shows at what stock to order. First decide which metric measures your success and only then look at the stock: different metrics have different stocks, and this is not a calculation error.
Common mistakes
Taking demand spread by months rather than by the protection period, ignoring lead-time spread, mixing service level with fill rate, taking standard deviation from a small sample. Measure the spread over 12 or more periods and check the units: a week in demand and a week in lead time.
What data is needed?
- Average demand and its weekly spread, lead time and its spread.
- The usual order size and the review period.
- Targets: the cycle service level and the fill rate.
Inputs
| Field | Unit | Value in the example | Meaning |
|---|---|---|---|
| Average weekly demand | units/week | 120 | The average over sales history without out-of-stock periods. The example value is illustrative. |
| Demand spread (std. deviation) | units/week | 30 | How far weekly demand departs from the average. For footwear and children’s count by size: the spread by size is higher than by model. |
| Lead time | weeks | 4 | From the order to receiving at the warehouse. |
| Lead-time spread | weeks | 0.5 | The standard deviation of the actual lead time over past deliveries; zero if the lead time never varies. |
| Review period | weeks | 1 | Zero: you order on stock level (reorder point). Above zero: you order on a schedule up to a target level. |
| Order size | units | 480 | How much you usually order at a time: needed for the fill rate. The larger the order, the less stock the same fill rate needs. |
| Target cycle service level | % | 95 | The probability of reaching the next delivery without a shortage. Your choice. The example value is illustrative. |
| Target fill rate | % | 98 | What share of demand must be met from the shelf. Your choice. The example value is illustrative. |
| Unit purchase price | ₽ | 1,200 | For the money tied in stock. The example value is illustrative. |
| Holding cost per year | % of price | 25 | Cost of money, storage, spoilage as a percentage of the price per year. There is no single standard. |
How it is calculated
| Step | Formula |
|---|---|
| Protection spread | √((lead + review) × demand σ² + demand² × lead σ²) |
| By service level | z(α) × protection spread |
| By fill rate | G(z) = (1 − β) × order ÷ spread; stock = z × spread; G(z) = φ(z) − z × (1 − Φ(z)) |
| Order-up-to level | demand × lead (+ review for scheduled ordering) + safety stock |
Worked numbers
Default values from the app; in the interactive version you enter your own.
- Demand spread over the protection period
- 90 unitsThe root of (lead + review) × demand spread squared + demand × lead spread squared.
- Safety stock by service level
- 148 units
- Safety stock by fill rate
- 78 units
- Difference: service level minus fill rate
- 70 unitsThat much extra you hold if you protect each cycle and not the share of demand.
- Fill rate at the service-level stock
- 99.6 %
- Service level at the fill-rate stock
- 80.7 %
- Order-up-to level: by fill rate
- 678 units
- Order-up-to level: by service level
- 748 units
- Expected shortage per order cycle (by fill rate)
- 9.6 units
- Money in safety stock (by fill rate)
- 93,627 ₽
- Holding of safety stock per year (by fill rate)
- 23,407 ₽
- By cycle service level
- By fill rate
Data as a table
| Target, % | By cycle service level | By fill rate |
|---|---|---|
| 80 | 75.7 | 0 |
| 90 | 115 | 0 |
| 95 | 148 | 27 |
| 97.5 | 176 | 66.6 |
| 99 | 209 | 110 |
| 99.5 | 232 | 139 |
| 99.9 | 278 | 195 |
Safety stock at different targets
| Target, % | By service level, units | By fill rate, units | Difference, units |
|---|---|---|---|
| 80 | 76 | 0 | 76 |
| 90 | 115 | 0 | 115 |
| 95 | 148 | 27 | 121 |
| 97.5 | 176 | 67 | 110 |
| 99 | 209 | 110 | 99 |
| 99.5 | 232 | 139 | 93 |
| 99.9 | 278 | 195 | 83 |
The normal distribution of weekly demand allows negative values: when the spread exceeds a third of the average the estimate is approximate. For footwear and children’s count by size, for accessories by article; seasonal one-batch goods are counted as “How much to order under uncertain demand”.
Nuances and limits
- Mixing the cycle service level with the fill rate: equal percents give different stock.
- Ignoring the lead-time spread: it is often more important than the demand spread.
- Counting by model when shortages occur by size: stock is needed by size.