How to Calculate Lead Time: A 7-Step Method

Calculating lead time is a data-handling procedure rather than an arithmetic problem. The subtraction itself is trivial; what decides whether the answer is usable is the sequence of choices around it — which two events you measure between, whether you count calendar or working days, how many orders you include, and what you add to the average before anyone plans against it. This page sets out a repeatable seven-step method, runs twelve real order records through it end to end, and shows why an average lead time of 13.7 days becomes a planning figure of 19.6 days once the spread is priced in. The step most often skipped is the last one, and it is the one that decides whether the number holds up.

Three Decisions Before You Start

Every disagreement about a lead time number traces back to one of these three, and all three are cheaper to settle now than to argue about later.

Which two events? Requisition to goods receipt, purchase order to goods receipt, and despatch to delivery are three different measurements of the same order. Pick one and write it down.

Which unit basis? Calendar days include weekends and holidays; working days exclude them. Both are valid and they produce different numbers from identical timestamps.

Which lead time? Actual, quoted and planned are three separate figures. Actual comes from the transaction record, quoted from the supplier, planned from your ERP master data. A gap between actual and planned is a master-data defect; a gap between actual and quoted is a service defect. This method calculates the actual.

The 7-Step Method

Step 1 — Fix the Start and End Trigger

Name one timestamp as the start and one as the end, and apply them to every record without exception. The most common failure here is measuring from the purchase order because that field is cleanest, while the business experiences the wait from the requisition.

That block is not small. The cross-industry median for requisition through purchase order release is 55 hours, roughly two and a half working days, with the slowest industries at 90 hours. Starting the clock at the PO makes procurement lead time look better than anyone experiences it.

Step 2 — Choose the Unit Basis

Working days for calculation, calendar dates for commitment. Working days measure capacity, which is what schedules consume. A customer, a production slot and a project plan all need a date.

Twenty working days and twenty calendar days differ by eight days on a five-day week, before any holiday. If different stages of your chain run on different working weeks — a six-day factory, a continuous port operation — convert stage by stage rather than applying one multiplier to the whole chain.

Step 3 — Pull One Record Per Order

One row per order, never a monthly summary. Each row needs the order identifier, the start timestamp, the end timestamp and the unit basis. Summarised data cannot produce a standard deviation, and without a standard deviation steps 6 and 7 are impossible.

Step 4 — Calculate Each Order's Lead Time

In Excel, =NETWORKDAYS(start, end, $H$2:$H$20)-1. The -1 is required because NETWORKDAYS counts both endpoints — an order placed and received on the same day returns 1 rather than 0. Omit it and every figure you report is one day long, consistently, which is why it so rarely gets noticed.

Step 5 — Calculate Average and Median

Calculate both and compare them. If the average sits above the median, the distribution is right-skewed — a small number of very late orders pulling the mean up. That is the normal shape for lead time, because an order can be a month late but can never be more than zero days early.

Step 6 — Measure the Spread

Use STDEV.S, not STDEV.P. Your order history is a sample of the supplier's ongoing behaviour, not the complete population, so the denominator is n − 1.

This is the number that costs money. Two suppliers can share an average of 14 days while one delivers between 12 and 16 and the other between 7 and 30. The second requires far more inventory to buy from, and nothing in the average says so.

Step 7 — Add a Buffer and Validate

Z is set by your target service level:

Then validate. The formula assumes a normal distribution and real lead times are right-skewed, so compare the calculated figure against the observed 95th percentile of your own data. If they diverge widely, trust the observed percentile and collect more records.

Worked Example: Twelve Orders End to End

A distributor pulls twelve receipts for one purchased component, measured requisition to goods receipt, in working days.

Running steps 5 through 7:

Three things worth reading off that table.

The average is not the answer. Planning against 13.7 days means missing roughly half of all orders. The figure to load into a system is 19.6.

The skew is visible. The average sits 1.2 days above the median because PO-2111 and PO-2112 stretch the tail. Ten of twelve orders landed between 10 and 15 days; two did not, and those two set the buffer.

The validation passes. The calculated 19.6 and the observed 20.3 are close enough to trust. Had the observed percentile come back at 28, the normal assumption would have been wrong and the raw percentile should win.

Calculating Lead Time With No History

When there are no records — a new supplier, a new part — build the figure from stages instead of from data.

Sum the stages, then treat the result as provisional and start recording actuals from the very first order. Within a quarter the estimate is replaced by measurement, and the estimate's only remaining job is to be compared against what actually happened.

Two cautions. A supplier's quote carries no spread, so a stage-built estimate cannot produce a standard deviation — meaning step 7 is unavailable and the figure has no buffer. And if any component sits on a long lead time, the chain is governed by that item rather than by the sum. A material lead time calculation takes the maximum across the bill of material, not the total.

Five Mistakes That Survive Review

Averaging quoted lead times instead of measuring actuals. Quotes describe intent. Only receipts describe behaviour.

Excluding the outliers. The two late orders in the worked example look like anomalies and are precisely what the buffer exists to cover. Removing them halves the standard deviation and produces a planning figure that fails exactly when it matters.

Mixing unit bases inside one dataset. Some rows in calendar days, some in working days, one column heading. The average is then meaningless and nothing about it looks wrong.

Using STDEV.P. Order history is a sample. STDEV.P treats it as the whole population and understates the spread, which understates the buffer.

Stopping at the average. The most common of the five. An average with no spread beside it cannot size a buffer, cannot compare two suppliers fairly, and cannot tell you whether the number is stable enough to commit to.

FAQs About Calculating Lead Time

How do you calculate lead time step by step?

Fix the start and end trigger, choose calendar or working days, pull one record per order, calculate each order’s elapsed time, then compute the average, median and standard deviation together. Finally add a buffer sized to your service level. On 12 orders averaging 13.7 working days with a 3.6-day standard deviation, the 95% planning figure is 19.6 days, not 13.7.

What is the average lead time formula?

Average Lead Time = Sum of All Order Lead Times ÷ Number of Orders. Twelve orders totalling 164 working days give an average of 13.7. Report the median alongside it — in this sample the median is 12.5 days, and the 1.2-day gap is caused by two late orders stretching the tail.

How many orders do I need before the number is trustworthy?

Twelve is enough to demonstrate the method and too few to trust a percentile. Aim for 30 or more before setting a buffer, and 100 or more before publishing a service commitment. Small samples produce unstable tails, and the tail is exactly what a buffer is sized against.

Should I use the average or the median lead time?

Use the average to calculate the buffer, because the standard deviation is measured around the mean. Use the median to understand typical experience. When the two diverge, the distribution is skewed — which is normal for lead time, since orders can be very late but never less than zero days early.

How do I calculate lead time when I have no history?

Build it from stages rather than from records. Break the chain into requisition, approval, supplier production, transit and receiving, estimate each from the supplier’s quote and your own process, and add them. Then treat the result as provisional and start recording actuals from the first order, so the estimate is replaced by measurement within a quarter.

Do I calculate lead time in calendar days or working days?

Calculate in working days and commit in calendar dates. Working days measure capacity, which is what a schedule needs. Calendar dates are what customers and production slots need. State the basis with every figure — 20 working days and 20 calendar days differ by 8 days on a five-day week before any holiday.

What buffer should I add to the average lead time?

Multiply the standard deviation by the Z-score for your target service level: 1.645 for 95%, 1.282 for 90%, 2.326 for 99%. On a 3.6-day standard deviation, a 95% service level needs a 6.0-day buffer. Validate it against your own history — the calculated buffer assumes a normal distribution and real lead times are usually right-skewed.

The Bottom Line

Calculating lead time takes seven steps, and six of them are about the data rather than the arithmetic. Fix the triggers, fix the unit basis, use order-level records, and report the average, median and standard deviation together. Then add a buffer sized to your service level and check it against your own observed percentile before anyone commits to it. The worked example above moves from an average of 13.7 working days to a planning figure of 19.6 — a 44% increase that comes entirely from pricing in the spread, and that is the difference between a number that describes the past and a number you can plan against.

Related Articles

• Lead Time Formula: Every Variant and When to Use Each — the six formulas behind this procedure.
• What Is Lead Time? Definition, Types, How to Measure — the ten types and which triggers each one uses.
• How Lead Time Variance Inflates Safety Stock — what the standard deviation from step 6 costs in buffer stock.
• How to Calculate Reorder Point When Lead Time Varies — turning these figures into a replenishment trigger.
• How to Reduce Lead Time: 23 Strategies — what to do once you know where the days are.

Related Calculators

• Lead Time Calculator — total order-to-delivery time across all stages.
• Lead Time Calculator in Excel — build steps 4 to 7 as a tracker.
• Lead Time Calculator Working Days — the unit basis from step 2.
• Procurement Lead Time Calculator — the requisition-to-receipt chain from step 1.
• Material Lead Time Calculator — when the bill of material governs instead.
• Lead Time Date Calculator — convert the planning figure into a promised date.
• All lead time calculators — the full set.

External references: APQC publishes the open-standard benchmark for purchase order cycle time referenced in step 1, and the Association for Supply Chain Management maintains the dictionary definitions of the lead time types this method measures.