What is safety stock? How suppliers buffer against retailer demand
Peter L. King's three formulas, the Z-score for each service level, a worked example in units, cycle service level versus fill rate, and what pushes the number up or down.
For a supplier to a UK retailer, it works twice. Where the retailer sizes its reserve with formulas like King's, it does so partly on the lead time it holds for you, and on how much that lead time varies. You keep a reserve of your own for the retailer's orders, sized the same way from your own figures.
Short answer
Safety stock is extra stock held so that most replenishment cycles end without a stockout, even when demand runs above plan or the goods arrive late. The formulas Peter L. King set out in APICS magazine in 2011 multiply a Z-score for the target service level by the variability in demand, in lead time, or in both.
What is safety stock?
King's definition is short: "Safety stock simply is inventory that is carried to prevent stockouts." He lists the causes it covers: "fluctuating customer demand, forecast inaccuracy, and variability in lead times for raw materials or manufacturing."
Wikipedia puts the supplier into the picture. Safety stock "acts as a buffer stock in case sales are greater than planned and/or the supplier is unable to deliver the additional units at the expected time."
It is not meant to stop every stockout. King writes that safety stock determinations "are not intended to eliminate all stockouts", only the majority of them. At a 95 percent service level, he expects the safety stock to go unused in about half of the cycles and to be enough in another 45 percent. "But in approximately 5 percent of replenishment cycles, expect a stockout."
So it is a trade-off. In King's words, "There must be a balance between inventory costs" and customer service. Safety stock also counts in the average stock behind a stock-turn figure: the worked example in the inventory turnover guide has the retailer keep 100 units of it throughout.
Where does safety stock sit in a reorder point?
It is the cushion on top of the stock expected to sell while an order is on its way.
Reorder point = demand over the lead time + safety stock
Wikipedia describes the first term as the figure that "is the average demand during the lead time." In the worked example below, 250 units of lead-time demand plus 91 units of safety stock gives a reorder point of 341 units.
The stock replenishment guide covers reorder points and the other ways a retailer reorders, and adds: "Where a retailer reorders this way, the lead time agreed with you is part of that calculation."
How do you calculate safety stock?
Pick the formula that matches what actually varies. All three below come from Peter L. King, "Crack the code: Understanding safety stock and mastering its equations", APICS magazine, July/August 2011. The symbols:
| Symbol | Meaning |
|---|---|
| Z | The Z-score for the target cycle service level (table below) |
| σD | Standard deviation of demand |
| σLT | Standard deviation of lead time |
| Davg | Average demand |
| PC | Performance cycle, King's term for total lead time |
| T1 | The time period σD is measured over |
King's performance cycle covers the time to decide what to order, communicate the order to the supplier, manufacture and process, deliver and store, and get back to the start of the next cycle. The lead time guide covers how lead time is set and measured.
1. Demand varies, lead time is reliable
Safety stock = Z × √(PC/T1) × σD
Use it when: the goods arrive on time and demand is the main uncertainty.
The units have to match. If σD is measured per day and the lead time is in days, PC/T1 is the lead time in days, which gives the shorthand: Safety stock = Z × σD × √(lead time). King's own example uses weekly demand data with an 8-day lead time: 1.65 × √(8/7) × 10 rolls a week, so "Safety stock = 18 rolls".
2. Lead time varies, demand is steady
Safety stock = Z × σLT × Davg
Use it when: demand is steady and, in King's words, "variability in lead time is the primary concern". Measure σLT in the same time unit as Davg's period. King's demand is weekly, so he converts half a day to about 0.07 weeks.
This is the term that carries a supplier's delivery reliability.
3. Both vary
Safety stock = Z × √((PC/T1 × σD²) + (σLT × Davg)²)
Use it when: demand and lead time both vary, from different causes, and both follow a normal distribution, as King's formula assumes. In King's words, "the safety stock is Z-score times the square root of the sum" of the squared individual terms, which gives a lower total than adding the two calculations together.
If the same cause drives both, King says "this equation can't be used". Add the two terms instead:
Safety stock = (Z × √(PC/T1) × σD) + (Z × σLT × Davg)
Z-scores by cycle service level
These values are from the table in King's article.
| Target cycle service level | Z-score |
|---|---|
| 90% | 1.28 |
| 95% | 1.65 |
| 97% | 1.88 |
| 98% | 2.05 |
| 99% | 2.33 |
With no safety stock at all, a Z of zero, King notes there is still enough stock "in 50 percent of cycles". With a Z of 1, "there will be enough inventory 84 percent of the time." The curve is not linear: "Higher cycle service levels require disproportionally higher Z-scores and, thus," disproportionately more safety stock. King puts typical goals between 90 and 98 percent, and treats a 100 percent cycle service level as statistically unattainable.
A worked example in units
The figures here are illustrative, not any retailer's real data. Average demand is 50 units a day, with a σD of 10 units a day. The lead time is 5 days, with a σLT of 1 day. The target cycle service level is 95%, so Z is 1.65. Demand is measured daily, so PC/T1 is 5.
| Step | Arithmetic | Result |
|---|---|---|
| Demand term only | 1.65 × 10 × √5 | 36.9 units |
| Lead-time term only | 1.65 × 1 × 50 | 82.5 units |
| Both, independent causes | 1.65 × √(5 × 10² + (1 × 50)²) = 1.65 × √3,000 = 1.65 × 54.77 | 90.4, rounded up to 91 units |
| Both, same cause (the sum) | 36.9 + 82.5 | 119.4, about 120 units |
| Reorder point | 50 × 5 = 250, plus 91 | 341 units |
Now change the supplier's side. If the lead time varies by half a day instead of a day (σLT 0.5), safety stock falls to 1.65 × √(500 + 625) = 55.3, rounded up to 56 units, and the reorder point to 306. If instead the lead time shortens to 4 days with σLT still at 1 day, safety stock is 88.9, rounded up to 89 units, and the reorder point is 200 + 89 = 289.
The service level matters too. With the same inputs, 90% (Z 1.28) gives 70.1 units before rounding, and 97% (Z 1.88) gives 103.0.
Rounding up to whole units is our convention. Here, lead-time variability is the larger term. That is the reverse of King's plastic film example, where he found the effect of demand variability "is almost 10 times that of lead time variability." Which term dominates depends on the numbers.
Where the formulas fall short
Wikipedia says: "No universal formula exists for safety stock". It names seasonal demand as one case: "For seasonal demand (for example high in summer, low in winter), the formula will consistently produce stock outs in summer and waste in winter."
Rules of thumb are weaker still. King notes that some businesses set safety stock as a share of cycle stock, "10 or 20 percent, for example. While easy to execute, such techniques generally result in poor performance." Rules built on maximum demand and maximum lead time are widely shared online, but neither King's article nor Wikipedia uses one.
Treat the result as a starting point. King's advice is to check whether the safety stock is being used in about half of the cycles and whether the service level target is being met, and to find the root cause before changing the number.
What does service level mean in a safety stock formula?
It means cycle service level: the share of replenishment cycles that end without a stockout. In a 2004 paper, Chopra, Reinhardt and Dada set it beside fill rate: "The fill rate measures the proportion of demand that is met from stock, whereas the cycle service level measures the proportion of replenishment cycles where a stockout does not occur."
King writes that fill rate is often a better measure of inventory performance, "as cycle service level merely indicates the frequency of stockouts without regard to their magnitudes." The Z table above is for cycle service level.
Cycle service level is also not the OTIF figure a retailer keeps on its suppliers. OTIF is a delivery measure defined by the customer's manual or agreement, and the OTIF guide covers it.
What makes safety stock go up or down?
The first two columns follow King's formulas. The third is our explanation of where a supplier has a hand in it.
| Input | If it rises, safety stock | What a supplier influences |
|---|---|---|
| Service level target | Rises, and disproportionately near the top | Nothing directly: the retailer sets its own target |
| Demand variability | Rises | Nothing directly: its own safety stock reflects the swings in retailer orders, which tend to be bigger |
| Lead time | Rises, with the square root of the lead time in the demand term | The lead time it agrees |
| Lead-time variability | Rises at King's typical 90 to 98% goals (see the caveat below) | How reliably it meets the due date |
| Average demand | Rises, through the lead-time term | None |
One caveat on the lead-time variability row. Chopra, Reinhardt and Dada found that between a 50 percent cycle service level and a threshold that lay between 50 and 70 percent in their tests, cutting lead-time variability can raise safety stock once it is worked out exactly rather than with the formula's normal approximation. They also argue that many firms operate in that range. At higher targets the formula's direction held: "For a cycle service level of 95%, however, the prescriptions of the normal approximation are correct".
There are also ways round holding it. King names an order-expediting process, finish-to-order, and make-to-order, which "eliminates the need for most safety stock" where lead times allow.
What does a retailer's safety stock mean for its suppliers?
Your lead time is one of its inputs. The lead time guide says: "Your customer holds the lead time you agreed, and the due date on each order comes from it." The retailer raises each purchase order "with a due date set from the lead time agreed with the supplier", as the stock replenishment guide puts it. The lead time guide also quotes the Lean Enterprise Institute: "safety stock protects you from incapability in your upstream processes and your suppliers."
So two of the inputs to a retailer's safety stock are yours to shape: the lead time it holds for you, and how much your actual deliveries vary around it. At a 95 percent target, less variability means less safety stock, as the half-day case in the example shows.
The messages around each order carry part of that picture. The purchase order guide explains: "An order response can acknowledge an order, accept it, propose changes or decline all or part of it, and each customer sets whether one is required." An accurate order response tells the buyer early what will and will not arrive. Then "the ASN bridges the gap between an order being placed and the goods physically arriving at a store or distribution centre", as the ASN guide puts it.
Your own demand tends to swing more than the retailer's. The bullwhip effect guide explains: "A supplier planning from purchase orders sees demand only after the retailer's forecasting, batching and promotions have shaped it, so its swings tend to be bigger." Each tier ordering upstream "will add safety stock to the amount actually needed." At peak, the peak season guide notes that "Safety stock buffers shrink faster than planned".
You hold your own safety stock too. King's worked example is one: a warehouse holding rolls of plastic film sold on to processors, sized from its weekly demand and the production and transport lead time behind it. For yours, σD comes from the retailer's orders and the lead time from your own production and inbound supply, and the history to measure both sits in your own systems. As the warehouse management system guide says, "The ERP records the customer's order as a sales order; the WMS runs the warehouse work that fulfils it, and the two share stock and despatch data."
How Transalis fits
The EDI page says: "Transalis now offers EDI in three categories."
- Web EDI: "A managed web portal to receive orders, send invoices and manage compliance in a browser."
- Integrated EDI: "EDI wired directly into your ERP, accounting or warehouse system."
- Enterprise EDI: "A scalable, enterprise-grade platform for high transaction volumes across thousands of trading partners, countries and systems."
Transalis also sends order responses and receiving advice messages for its customers. The web EDI guide covers the browser route: "Web EDI lets a supplier exchange EDI documents through a web browser, with no EDI software or IT team."
Book a 30-minute call with a Retail Supply Chain specialist, who will walk you through EDI and eInvoicing and how Transalis makes this simple and easy.
Frequently asked questions
- What is the difference between safety stock and buffer stock?
Usage varies. Wikipedia describes safety stock as acting as a buffer stock in case sales run above plan or the supplier delivers late, and many people use the two as one term. Some use buffer stock for a wider reserve held for events or seasons. No authoritative source we read settles the split.
- What is the difference between minimum stock and safety stock?
Where minimum stock means the trigger level in a min/max reorder rule, it sits above the safety stock, because it also covers the demand expected while the next order arrives. Safety stock is the part of that level held for variability in demand and lead time. The stock replenishment guide on transalis.com covers min/max as a method.
- What is King's formula for safety stock?
Peter L. King gave three in APICS magazine in 2011. Demand varies: Z × √(PC/T1) × σD. Lead time varies: Z × σLT × Davg. Both vary independently: Z × √((PC/T1 × σD²) + (σLT × Davg)²). Z comes from the target cycle service level, for example 1.65 at 95 percent.
- What is the 50% rule for safety stock?
No source we read defines a 50 percent rule. King warns against setting safety stock as a share of cycle stock, 10 or 20 percent for example, because such techniques generally result in poor performance. A formula built on your own demand and lead-time data, checked against real stockouts, is the sounder route.
- Can you give an example of safety stock?
Take illustrative figures: demand of 50 units a day with a standard deviation of 10, and a 5-day lead time with a standard deviation of 1 day. At a 95 percent cycle service level, Z is 1.65 and the combined formula gives about 91 units of safety stock, for a reorder point of 341 units.
- How much safety stock should a supplier hold?
There is no single figure. It depends on how much the supplier's own demand from retailer orders varies, on its production and inbound lead times and how much they vary, on the lead times its customers allow, and on the service target it sets. Make-to-order removes most of it where those lead times allow.
Read next
- 13 October 2021EDI in the supply chainLearn how EDI in the supply chain allows businesses, from SMEs to large corporations, to automatically exchange data with eachother…Read more
- 13 October 2021eDI Instant: how to get started with EDI todayHow to give your business a boost and get started with EDI. The latest product from supply chain technology provider Transalis, eDI Instant.Read more
- 17 December 2020Getting Started with EDI?How to get started with EDI: choose Web, Integrated or Enterprise EDI from Transalis, connect your trading partners and start exchanging…Read more