Lead time variance is a statistical dispersion measure that enters the safety stock formula multiplied by average daily demand, and that multiplication is why it dominates the result. Under one standard input set — 100 units a day, a demand standard deviation of 20 units, a 20-day (4-week) average lead time, a lead time standard deviation of 5 days — the lead time term holds 96.9% of the value inside the radical and the demand term holds 3.1%. Lead time variance drives the buffer harder than average lead time does. This page gives you the combined formula, a worked comparison of two reduction scenarios, two sensitivity tables, and a dominance test to run against your own records. Five inputs set the answer: the Z-score, average lead time, average daily demand, the demand standard deviation, and the lead time standard deviation. Cutting the lead time standard deviation from 5 days to 2 days removes 476 units and $5,712 per stock keeping unit (SKU) at an unchanged 95% cycle service level.
On this page
What Is Lead Time Variance?
Lead time variance is the statistical dispersion of observed replenishment durations around their average, and it works as an inventory policy parameter rather than a performance score. Two suppliers quoting the same 20-day (4-week) average lead time can hold completely different variance, and the enterprise resource planning (ERP) record usually stores only the average, so the variance-driven buffer expansion it causes stays invisible until a stockout exposes it.
Three measures describe that dispersion. Lead time variance is the average squared deviation from the mean, in squared days. Lead time standard deviation is its square root, in days. The coefficient of variation is the standard deviation divided by the average, a unitless ratio: a 5-day standard deviation against a 20-day average gives 0.25, against a demand coefficient of variation of 0.20 on the same item.
The word "variance" arrives carrying five meanings, and a planning conversation stalls when two people use two of them. Select each sense below to see what it measures and what it costs you.
Sense 1 — statistical variance, σ²
The average squared deviation from the mean, measured in squared days. A 5-day lead time standard deviation is a lead time variance of 25 squared days. The safety stock formula never takes this number directly.
Sense 2 — standard deviation, called variance in daily speech
The square root of sense 1, measured in days. This is the number the formula takes, and it is what most operations teams mean when they say "lead time variance". Senses 1 and 2 differ by a square root, which is a factor of 5 on these inputs.
Sense 3 — budget variance
Actual against plan, measured in currency. A finance reader arrives holding this sense by default, and it has no place in the safety stock formula. Name it once, then set it aside for the rest of any planning meeting.
Sense 4 — supplier variance, quoted against actual
The gap between the quoted lead time and the delivered lead time. This is a mean shift, not a dispersion, and no quantity of safety stock corrects it. Fix it by renegotiating the quoted duration or by updating the planned lead time parameter in the material requirements planning (MRP) record.
Sense 5 — a permitted variance
An exception granted against a standard, as in a legal or zoning variance. It shares a spelling with the statistical quantity and nothing else. It appears here only so that it stops appearing in your search results.
Key Takeaways
- Lead time variance holds 96.9% of the value inside the safety stock radical on these inputs. The demand term holds 3.1%.
- Cutting average lead time by 25%, from 20 days to 15 days, frees 3 units of safety stock.
- Cutting lead time deviation by 3 days, from 5 days to 2 days, frees 476 units — a 57% reduction worth $5,712 per SKU at $12 per unit.
- The formula is
SS = Z × √(LT × σD² + D² × σLT²). The square root encloses both terms. - Run the dominance test: compare LT × σD² against D² × σLT². Here the lead time term is 31 times the demand term.
- Measure on 20 to 30 receipts per supplier-SKU pair, strip special-cause outliers first, and check the result against the observed 95th percentile.
Variance, Standard Deviation, and Variability Are Not the Same Term
Variance, standard deviation, and variability are three separate things, and this page keeps them separate throughout. Lead time variance is the squared quantity, σLT², measured in squared days. Lead time standard deviation is σLT, measured in days. Variability is the general condition of a lead time that moves, with no unit attached.
On the inputs used here, the lead time standard deviation is 5 days and the lead time variance is 25 squared days. Feeding 25 into the σLT slot instead of 5 returns 4,128 units of safety stock instead of 838 — a five-fold error produced by one unit confusion.
Budget variance is a fourth thing entirely. It measures actual spending against planned spending in currency, it belongs to the finance function, and it never enters an inventory calculation. Every use of "variance" below means lead time variance in the statistical sense.
Why Variance Drives Safety Stock Harder Than Average Lead Time
Lead time variance drives safety stock harder than average lead time because of what each input is multiplied by inside the radical. Average lead time multiplies the demand variance, 400 squared units. Lead time variance multiplies the square of average daily demand, 10,000 squared units. That is a 25-fold difference in coefficient before either input moves.
The demand term, LT × σD², comes to 20 × 400 = 8,000. The lead time term, D² × σLT², comes to 100² × 5² = 250,000. The lead time term is 31 times the demand term and holds 96.9% of the 258,000 total.
Volume is what makes lead time uncertainty costs land where they do. Average daily demand enters the lead time term squared, so a high-volume SKU pays for supplier unpredictability far more than a low-volume SKU holding the identical delivery record.
The Safety Stock Formula With Variable Lead Time
The safety stock formula with variable lead time combines two independent sources of uncertainty under one square root. Define the five symbols before using them: Z is the service factor read from the normal distribution for your chosen cycle service level, LT is average lead time in days, σD is the standard deviation of daily demand in units, D is average daily demand in units, and σLT is the standard deviation of lead time in days.
The Simple Formula and Where It Breaks
The simple formula assumes a fixed lead time and multiplies demand deviation by the square root of the duration:
Safety Stock = Z × σD × √LT= 1.65 × 20 × √20= 1.65 × 20 × 4.47 = 148 unitsThat form is correct only when lead time is deterministic. It returns 148 units (1.5 days of cover) against a correct answer of 838 units (8.4 days of cover), understating the buffer by 690 units, or 82.3%. Use it where a fixed lead time is contractually guaranteed and measured, and nowhere else.
The Combined Formula for Demand and Lead Time Variability
The combined formula adds a lead time term inside the same radical:
Safety Stock = Z × √( LT × σD² + D² × σLT² )= 1.65 × √( 20 × 20² + 100² × 5² )= 1.65 × √( 8,000 + 250,000 )= 1.65 × 507.94 = 838 unitsPeter L. King set out this combined form for a planning audience in APICS Magazine, published by the body now named the Association for Supply Chain Management (ASCM). The derivation assumes demand and lead time are independent and roughly Gaussian.
Check the square root before you use any version of this formula. Several copies circulating online lost the radical to bad PDF extraction and render as SS = Z × LT × σD² + D² × σLT². That form returns 425,700 units on these inputs — 508 times the correct answer. The radical encloses both terms, always.
What Each Term Contributes
| Formula part | Symbol | Unit | Value here | Share of the radical |
|---|---|---|---|---|
| Average lead time | LT | days | 20 (4 weeks) | multiplier on the demand term |
| Demand standard deviation | σD | units per day | 20 | squared to 400 |
| Demand term | LT × σD² | squared units | 8,000 | 3.1% |
| Average daily demand | D | units per day | 100 | squared to 10,000 |
| Lead time standard deviation | σLT | days | 5 | squared to 25 |
| Lead time term | D² × σLT² | squared units | 250,000 | 96.9% |
| Sum inside the radical | — | squared units | 258,000 | 100% |
| Square root | √ | units | 507.94 | — |
| Service factor | Z | unitless | 1.65 | linear multiplier |
| Safety stock | SS | units | 838 | 8.4 days of cover |
Selecting the Z-Score for Your Service Level
To select the Z-score, decide the cycle service level first, then read the service factor from the normal distribution. Cycle service level is the probability of avoiding a stockout during one replenishment cycle. Fill rate is a different measure — the fraction of demand served from stock on hand — and the two return different numbers on the same item. The Z-score in this formula is the cycle service level input.
In Microsoft Excel, =NORM.S.INV(0.95) returns 1.6449, rounded to 1.65 throughout this page. Use =STDEV.S(range) for both deviations, not =STDEV.P(range): your receipt history is a sample of the supplier's future performance rather than the full data set, so the sample function divides by n − 1 and returns the larger, correct estimate.
| Cycle service level | Z | Safety stock | Days of cover | Value at $12 |
|---|---|---|---|---|
| 90% | 1.28 | 650 units | 6.5 days | $7,800 |
| 95% | 1.65 | 838 units | 8.4 days | $10,056 |
| 99% | 2.33 | 1,183 units | 11.8 days | $14,196 |
| 99.9% | 3.09 | 1,570 units | 15.7 days | $18,840 |
Worked Example: What One Day of Variance Costs
One scenario runs through the rest of this page. A distributor stocks one SKU at $12 per unit, sells 100 units a day, and replenishes from a single overseas supplier.
The Baseline Scenario
The baseline carries average daily demand of 100 units, a demand standard deviation of 20 units, an average lead time of 20 days (4 weeks, 480 hours), a lead time standard deviation of 5 days, and a 95% cycle service level. Safety stock is 838 units, which is 8.4 days of cover and $10,056 of working capital in that one item.
Scenario A — Cut Average Lead Time by 5 Days
Scenario A moves average lead time from 20 days to 15 days, a 25% reduction, and holds everything else. The demand term falls from 8,000 to 6,000. The lead time term does not move, because average lead time does not appear in it.
= 1.65 × √( 15 × 400 + 10,000 × 25 )= 1.65 × √( 6,000 + 250,000 ) = 1.65 × 505.96= 835 units — a saving of 3 units, or $36A quarter off the average lead time frees 3 units of safety stock. That reduction is worth having for other reasons — 500 fewer units of pipeline inventory and a 5-day shorter planning horizon — and it is not a safety stock lever.
Scenario B — Cut Lead Time Deviation by 3 Days
Scenario B holds average lead time at 20 days and moves the lead time standard deviation from 5 days to 2 days. The lead time term falls from 250,000 to 40,000.
= 1.65 × √( 20 × 400 + 10,000 × 4 )= 1.65 × √( 8,000 + 40,000 ) = 1.65 × 219.09 = 361.5= 362 units — a saving of 476 units, or $5,712Three days off the lead time standard deviation frees 476 units, a 57% cut, while the average lead time and the 95% cycle service level both stay exactly where they were.
Comparing Both Scenarios in Units and Cash
| Measure | Baseline | Scenario A | Scenario B |
|---|---|---|---|
| Average lead time | 20 days | 15 days | 20 days |
| Lead time deviation | 5 days | 5 days | 2 days |
| Demand term | 8,000 | 6,000 | 8,000 |
| Lead time term | 250,000 | 250,000 | 40,000 |
| Safety stock | 838 units | 835 units | 362 units |
| Days of cover | 8.4 days | 8.4 days | 3.6 days |
| Inventory value | $10,056 | $10,020 | $4,344 |
| Units freed | — | 3 | 476 |
| Cash released | — | $36 | $5,712 |
Both scenarios cost real effort. Scenario A needs a faster route or a closer supplier. Scenario B needs delivery consistency from the supplier you already use. The arithmetic says the second one pays 159 times better on this buffer.
Measure your own stage-by-stage baseline first. Enter each stage of your replenishment to get a dated arrival estimate and a segment breakdown you can take a standard deviation from.
Open the calculator →Sensitivity: How Safety Stock Responds to Each Input
Safety stock responds to the five inputs at four different rates, and a safety stock sensitivity analysis is what tells you which rate applies to the input you are about to move.
| σLT (days) | Safety stock (units) | Days of cover | Change vs σLT = 0 |
|---|---|---|---|
| 0 | 148 | 1.5 | baseline |
| 1 | 221 | 2.2 | +49% |
| 2 | 362 | 3.6 | +145% |
| 3 | 517 | 5.2 | +249% |
| 5 | 838 | 8.4 | +466% |
| 8 | 1,328 | 13.3 | +797% |
| 10 | 1,657 | 16.6 | +1,020% |
One widely repeated claim states that moving from ±2 days to ±8 days "nearly doubles" safety stock. That claim is understated. On these inputs, a 2-day deviation gives 362 units and an 8-day deviation gives 1,328 units, a 3.7-fold increase. Run the arithmetic rather than repeating the claim.
Square Root Scaling on Average Lead Time
Average lead time scales under the square root and touches only the demand term, which is why it moves the answer so little. Doubling average lead time from 20 days to 40 days (4 weeks to 8 weeks) moves safety stock from 838 units to 851 units, a rise of 1.6%.
Linear Scaling on Lead Time Deviation
Lead time deviation scales linearly against demand volume, because D² × σLT² takes the square root back off the deviation. Doubling the lead time standard deviation from 5 days to 10 days moves safety stock from 838 units to 1,657 units, a rise of 98%. The same percentage move on the average produced 1.6%, and on the deviation it produces 98%.
The Demand-Squared Multiplier
Average daily demand enters the lead time term squared, so the cost of supplier unpredictability rises with volume. At 200 units a day with the same 20-unit demand deviation and the same 5-day lead time standard deviation, safety stock is 1,657 units against 838 units at 100 units a day. Days of cover barely moves, from 8.4 days to 8.3 days, and the unit count and the cash both double. Rank your variance reduction work by SKU volume, not by SKU count.
Service Level Escalation Above 95%
Safety stock scales linearly with the Z-score, so the unit cost of each extra point of cycle service level climbs as the points get smaller.
Moving from 95% to 99% adds 345 units and $4,140. Moving from 99% to 99.9% adds 387 units and $4,644 for nine-tenths of one percentage point. The 4 points from 95% to 99% cost 86 units per point; the 0.9 points from 99% to 99.9% cost 430 units per point, five times as much.
Lead Time Variance vs Demand Variance: Which Dominates?
To settle which variance dominates, compute both terms and compare them directly. The larger term is where improvement effort belongs, and the comparison takes one line of arithmetic:
Demand term = LT × σD² = 20 × 400 = 8,000Lead time term = D² × σLT² = 10,000 × 25 = 250,000Ratio = 250,000 ÷ 8,000 = 31× in favour of lead timeRun this test per SKU before starting any buffer reduction project. Demand variance takes over only when the lead time standard deviation drops below 0.89 days (21 hours) on these inputs — a level of delivery consistency that almost no external supplier reaches. Demand forecasting work improves the 3.1% term. Supplier consistency work improves the 96.9% term. Reversing that order is stockout risk mispricing: effort goes to the term that cannot move the answer.
How Variance Flows Into the Reorder Point
Lead time variance reaches the reorder point through the safety stock term and nowhere else. The reorder point adds expected demand during the lead time to the buffer:
Reorder Point = (D × LT) + Safety Stock= (100 × 20) + 838 = 2,000 + 838 = 2,838 unitsScenario B: (100 × 20) + 362 = 2,362 unitsSafety stock holds 29.5% of the baseline reorder point and 15.3% of it after the deviation cut. Lead time demand of 2,000 units does not move, because average lead time did not move. A planner reviewing a reorder point that looks too high should split it into these two parts first: the lead time demand half is arithmetic, and the safety stock half is where the variance sits.
Where Lead Time Variance Comes From
Lead time variance accumulates across five sources, and they add as variances rather than as standard deviations. A stable lead time at one stage does not protect the total when another stage runs wide.
Supplier Delivery Variability
Supplier lead time variance comes from production scheduling at the supplier, raw material availability, and order priority. A supplier running your order against a level schedule delivers a tight distribution; a supplier slotting your order wherever capacity appears delivers a wide one. Both report the same average, and the ERP record stores the same planned duration for each.
Transit and Customs Variability
Transit time variance and customs clearance variance sit outside your operation entirely. Ocean routes carry port congestion, vessel roll-overs and weather, so a nominally deterministic 28-day sailing arrives across a 10-day window. Customs clearance variance is bimodal rather than continuous: most shipments clear in hours, and an inspected shipment adds 3 to 10 days.
Internal Queue and Approval Variability
Queue time variance and approval time variance are self-inflicted and the cheapest to remove. A requisition waiting for a signature that arrives on Monday or on Thursday adds 3 days of spread for no reason connected to supply. Internal variance carries no supplier negotiation and no capital cost, which is why it belongs first in any reduction sequence.
Batch Release and Shipment Consolidation Effects
Consolidation creates variance out of an otherwise predictable replenishment. Holding orders until a container fills means an order placed the day after a sailing waits a full cycle, so the same reliable lane produces a wide distribution at the receiving end. The fix is a fixed sailing calendar, which converts a random wait into a uniform lead time with a known ceiling.
Seasonal and Capacity-Driven Variance
Seasonal lead time variance appears when supplier capacity tightens against demand: Lunar New Year, peak shipping season, and quarter-end order surges all stretch the upper tail without moving the average much. Compute seasonal segments separately. Pooling a stable 8 months with a congested 4 months produces a standard deviation that is too high for most of the year and too low for the peak.
Four conditions describe the opposite end of each source: a deterministic lead time with zero variance, a reliable delivery against a fixed schedule, predictable replenishment on a uniform lead time, and a tight distribution with a short upper tail. Those conditions are what a variance reduction programme is buying, and reaching them is what stops the buffer stock overadjustment that follows every late delivery.
How to Measure Lead Time Variance
To measure lead time variance, build an order-level data set, compute the mean and standard deviation on it, then test the distribution before trusting the formula's output.
Build the Order-Level Data Set
Pull four fields per order: purchase order number, order date, goods receipt date, and supplier. Elapsed days is the receipt date minus the order date. Use the goods receipt date rather than the ship notification date, because the buffer protects the interval until stock is available to pick.
Calculate the Mean, Variance, and Standard Deviation
Use =AVERAGE(range) for the average lead time and =STDEV.S(range) for the standard deviation, in days. =VAR.S(range) returns the variance in squared days if you want to see both. Behind those two functions sit the squared deviations from the mean, their sum of squares, and a division by the degrees of freedom, n − 1. The formula takes the standard deviation, so read the output of STDEV.S into the σLT slot.
Test the Distribution Before Trusting the Formula
The combined formula assumes a roughly normal distribution of lead times. Real supplier lead times are right-skewed: delivery arrives arbitrarily late and never earlier than zero, so lead time distribution tails run long on the upper side and the formula understates them.
These 40 receipts average 20.0 days with a standard deviation of 5.0 days, matching the inputs used throughout this page, and the median sits at 18.5 days. The average above the median is the signature of a right-skewed distribution. The normal distribution places the 95th percentile at 28.3 days, while the observed 95th percentile sits at 29 days — 0.75 days (18 hours), or 75 units of demand, beyond what the formula protects.
Compare the calculated safety stock against the observed 95th percentile of lead time demand every time you size a buffer. Use percentile-based sizing, if the two figures diverge by more than a few percent. A Gamma distribution fits right-skewed lead times better than the Gaussian, and Monte Carlo simulation over the actual receipt history skips the distributional assumption entirely.
Segment by SKU, Supplier, and Lane
Segment before sizing. One catalog-wide lead time standard deviation averages a reliable supplier and an erratic one into a middling figure that overbuffers the first and underbuffers the second at the same time. Compute σLT per supplier-SKU pair, and per lane where one SKU arrives on more than one route.
Separate Common-Cause From Special-Cause Variation
Walter A. Shewhart drew the distinction between common-cause variation, which the process produces routinely, and special-cause variation, which comes from an identifiable event. W. Edwards Deming carried it into management practice, and it underpins Statistical Process Control (SPC) and Six Sigma.
One customs seizure, one plant fire, or one port strike belongs in contingency planning and dual sourcing, not in σLT. Averaging a 90-day disruption into a supplier's record inflates the buffer permanently against an event that will not repeat in that form. Strip identified special causes, size the buffer on common-cause variation, and log the stripped records separately so the decision stays auditable.
Two data conditions decide whether the number is usable. Sample size comes first: 20 to 30 receipts per supplier-SKU pair is a working minimum, and a standard deviation computed from 5 receipts moves by days when the sixth lands. Stationarity comes second: a supplier whose lead time drifted from 20 days to 30 days across a year has a trend, not variance, and pooling both halves reports a spread that no longer exists.
How to Reduce Lead Time Variance
To reduce lead time variance, attack the spread rather than the average, and start with the sources you control outright. Lead time variance reduction belongs in the reorder point system review alongside the sales and operations planning (S&OP) cycle, because the buffer it sets is a working capital decision as much as a service level one.
| Tactic | What it narrows | Typical σLT reduction | Cost |
|---|---|---|---|
| Fixed order and sailing calendar | Consolidation wait | 1–3 days | none |
| Removing approval queues | Approval time variance | 0.5–2 days | none |
| Delivery-window SLA with measurement | Supplier scheduling spread | 1–3 days | low |
| Rolling forecast sharing | Supplier material wait | 1–2 days | low |
| Customs pre-clearance and a broker SLA | Clearance variance | 1–4 days | low |
| Dual sourcing on critical parts | The upper tail only | 2–6 days at the tail | medium |
| Nearshoring | Transit and clearance together | 3–8 days | high |
Measure before and after on the same four figures: average, standard deviation, 95th percentile, and sample size. A tactic that moves the average while the standard deviation holds steady has not reduced the buffer at all. Reducing average lead time and reducing lead time deviation are two different projects with two different payoffs, and a project charter that says "reduce lead time" without naming which one produces the 3-unit result rather than the 476-unit result. The 23 lead time reduction strategies cover the average; the table above covers the spread.
The Bullwhip Connection
Lead time variance and the bullwhip effect amplify each other across the supply chain. Jay Forrester described the amplification of demand swings up a multi-stage chain in the 1960s, which is why it is also called the Forrester effect. Hau Lee named it the bullwhip effect and identified four operational causes: demand signal processing, order batching, price fluctuation, and rationing under shortage.
Longer and less predictable lead times worsen every one of those four causes, because each echelon forecasts further ahead and orders in larger batches to cover the uncertainty. Larger batches at one echelon produce lumpier demand at the next, which stretches its lead times in turn — replenishment delay propagation that the trade calls supply chain nervousness. Ford W. Harris established the batch-size arithmetic behind that behaviour with the Economic Order Quantity in 1913, and Little's Law fixes the relationship between work in progress and flow time at each stage.
Multi-echelon inventory optimization (MEIO) sizes buffers across the whole network rather than at each node separately, and it is the standard response where lead time variance compounds across three or more echelons. Single-echelon sizing, which is what the formula on this page performs, treats each node as independent and overbuffers a deep network.
Mistakes That Hide or Worsen Variance
Six practices hide lead time variance or make it worse, and five of them happen inside your own operation.
- Padding the planned lead time parameter. A planner who extends the MRP lead time "to be safe" moves the mean instead of narrowing the spread, releasing purchase orders earlier and growing work in progress while the buffer stays the same size.
- Storing one lead time per supplier. SAP, Oracle NetSuite and most ERP systems hold a single planned duration per material-vendor record. That field cannot represent a distribution, so the variance lives outside the system unless you compute it separately.
- Running an expediting culture. Expediting one order pushes every other order back in the same supplier queue, which widens the distribution for the whole book while appearing to fix one line.
- Averaging special causes into σLT. One port strike inside the sample permanently inflates the buffer against an event that will not repeat in that form.
- Setting safety stock by days-of-cover habit. A flat "two weeks of cover" rule ignores both variances and both volumes, and that inventory policy distortion is the most common source of excess working capital in a catalog.
- Reporting the average only. A lower average with an unchanged standard deviation produces no service level gain, because the buffer is sized by the spread.
That last mistake has a structural cause: "safety stock" names four different numbers inside most companies, and nobody reconciles them.
Sense 1 — the calculated buffer
What the combined formula returns from the five inputs: 838 units at a 95% cycle service level. This is the only one of the four numbers with a service level attached to it.
Sense 2 — the parameter in the ERP record
The safety stock field stored against the material master in SAP, Oracle NetSuite or your MRP system. It drives replenishment whether or not it matches sense 1, and it is usually the last number anyone reviews.
Sense 3 — the physical inventory
What is actually on the warehouse floor above the cycle stock. It diverges from sense 2 through receipt timing, returns, and quality holds, and only a cycle count reveals by how much.
Sense 4 — the days-of-cover rule of thumb
"Two weeks of cover", set by habit and carried forward for years. On these inputs that rule stores 1,400 units where the arithmetic asks for 838 — 562 units and $6,744 of excess working capital in a single SKU. The gap between sense 1 and sense 4 is where most excess inventory hides.
FAQs About Lead Time Variance and Safety Stock
How does lead time variance affect safety stock?
Lead time variance enters the safety stock formula as D² × σLT², so it is multiplied by the square of average daily demand. At 100 units a day with a 5-day lead time standard deviation, that term is 250,000 against 8,000 for the demand term, or 96.9% of the value inside the radical. Safety stock of 838 units falls to 362 units when the lead time standard deviation falls from 5 days to 2 days, with every other input unchanged.
What is the safety stock formula with variable lead time?
Safety Stock = Z × √(LT × σD² + D² × σLT²). Z is the service factor for the cycle service level, LT is average lead time in days, σD is the standard deviation of daily demand, D is average daily demand, and σLT is the standard deviation of lead time in days. The square root must enclose both terms. Several versions circulating online drop the radical through bad PDF extraction, which returns 425,700 units instead of 838 on these inputs.
Is lead time variance more important than demand variance?
Usually yes, and you can test it in one line. Compare LT × σD² against D² × σLT², and the larger term is where improvement effort belongs. On the inputs used here the lead time term is 31 times the demand term. The demand term only takes over below a lead time standard deviation of 0.89 days, which almost no supplier achieves.
Does reducing average lead time reduce safety stock?
Yes, but barely. Average lead time sits under the square root and multiplies only the demand term, so cutting it from 20 days to 15 days, a 25% reduction, moves safety stock from 838 units to 835 units. That is 3 units, worth $36 at $12 per unit. A shorter average lead time reduces pipeline inventory and shortens the planning horizon, which are real gains, and it is not a safety stock lever.
How many receipts do you need to calculate lead time standard deviation?
Use 20 to 30 receipts per supplier-SKU pair as a working minimum, and 40 or more where the lead time is long. A standard deviation computed from 5 receipts is noise, and it moves by several days when the sixth receipt lands. Below 20 receipts, pool the records at the supplier or lane level, size the buffer on the pooled figure, and switch to the SKU-level figure once the count is there.
What is the difference between variance and standard deviation in this formula?
Variance is the standard deviation squared, so the two differ by a square root and carry different units. A lead time standard deviation of 5 days is a lead time variance of 25 squared days. The formula takes the standard deviation, not the variance, even though the inner term squares it again. Feeding 25 into the σLT slot instead of 5 returns 4,128 units instead of 838.
Should safety stock be calculated per supplier or across the catalog?
Per supplier-SKU pair, and per lane where one SKU arrives on more than one route. A single catalog-wide lead time standard deviation averages a reliable supplier and an erratic one into one middling figure, which overbuffers the reliable supplier and underbuffers the erratic one at the same time. Segment first, then size.
How often should lead time variance be recalculated?
Quarterly for most categories, monthly where lead times are moving. Test for drift before pooling 12 months of records: a supplier whose lead time moved from 20 days to 30 days over a year has a trend, not variance, and averaging the two halves inflates the standard deviation with movement that already finished.
The Bottom Line
Lead time variance holds 96.9% of the value inside the safety stock radical on these five inputs, against 3.1% for the demand term, which is why it drives the buffer harder than average lead time does. Cutting average lead time by a quarter, from 20 days to 15 days, frees 3 units. Cutting the lead time standard deviation from 5 days to 2 days frees 476 units and $5,712 per SKU at the same 95% cycle service level. Measure that deviation on 20 to 30 receipts per supplier-SKU pair with STDEV.S, strip special-cause events first, and check the answer against the observed 95th percentile, because real lead times are right-skewed and the formula understates the upper tail. Measuring lead time deviation per supplier and per SKU is the highest-return change available to most planning teams.
Related Articles
- How to Calculate Reorder Point When Lead Time Varies — the same 838-unit buffer inside the full 2,838-unit replenishment trigger.
- What Is Lead Time? Definition, Types, How to Measure — the definitional groundwork behind σLT.
- How to Reduce Lead Time: 23 Strategies — tactics for the average, with days recoverable and cost for each.
- Lead Time Calculator — model a replenishment stage by stage.
- Supply Chain Lead Time Calculator — end-to-end multi-stage timelines.
- Procurement Lead Time Calculator — requisition to purchase order issue.
- Sea Freight Lead Time Calculator — port-to-port transit and clearance.
- Lead Time Calculator in Excel — build the receipt data set
STDEV.Sreads from. - All lead time calculators — the full set of 21 tools.