Lead Time: Definition, Formula, Little's Law & How to Cut It

What is lead time?

Lead time is the elapsed clock time between the moment a demand signal arrives and the moment the corresponding deliverable is ready. In manufacturing, it is the single most honest measure of how a value stream actually behaves under real conditions. Unlike cycle time or takt time, it cannot be improved on paper — only by changing what physically happens between order entry and shipment.

The definition sounds trivial and is almost always misapplied in practice. Three variants circulate in plants, used interchangeably and meaning very different things: the time the customer sees (customer lead time), the time production owns (production lead time) and the time a specific order spends in the shop (order lead time). Every KPI discussion about "reducing lead time" that does not first pin down which variant is meant ends in the wrong investment.

The lever matters because lead time compresses every other operational problem into one number. Long lead times signal high work-in-process (WIP), unstable processes, long changeovers, poor quality loops or excess buffering. Short, stable lead times signal flow. A modern MES exists mainly to make this number visible, traceable and improvable in real time rather than at month-end.

The three lead times that get confused

Before optimising anything, fix the definition. Each variant has a different owner, different root causes and different improvement levers.

In most mid-sized plants, production lead time is 2–5× longer than order lead time — because "released but not started" consumes days that nobody measures. Fixing that gap is usually a larger win than optimising the actual machining.

Lead time vs. cycle time vs. takt time

These three get swapped constantly, including in ISO 22400 discussions. They measure different things and move in different directions when the process improves.

In a healthy line, cycle time ≤ takt time, and lead time stays stable regardless of order size. The moment cycle time creeps past takt time, WIP grows and lead time explodes non-linearly — this is Little's Law in action.

Little's Law: the only formula that matters

Every lead-time discussion eventually reduces to one equation. Little's Law, proven in 1961, holds for any stable system regardless of product or industry:

Lead Time = WIP ÷ Throughput

The implication is brutal and counterintuitive: if you want shorter lead time, you cannot only push for faster machines. You must either reduce WIP or increase throughput. In most plants, reducing WIP is an order of magnitude easier — and it is exactly what pull control, Kanban and CONWIP loops are designed to do.

The faster machines buy 13 %. Cutting WIP buys 75 %. The maths is the same in every plant I have walked through in 25 years.

The anatomy of a real lead time: value-added vs. waste

The most sobering number in Lean audits is the ratio of value-adding time to total lead time. In typical metal, plastic and assembly operations, the part is actually being worked on for 2–10 % of its total time in the system. The other 90+ % is queue, transport, wait, inspection and rework. Cutting lead time means cutting that 90 %, not squeezing the already-tight 10 %.

How to actually reduce lead time

The levers are not new, and they are not secret. What fails most plants is sequencing: applying them in the wrong order, or all at once.

The common mistake is jumping to step 6 because it is visible and capex-friendly. Automating a process with 20 % WIP-driven waste locks the waste into faster hardware.

How an MES turns lead time from a guess into a live number

Paper-based shopfloors calculate lead time monthly, from ERP booking timestamps. The number is late, aggregated, and hides the variance that actually matters. A cloud MES changes three things: it captures the real clock at each operation, it exposes the queue time between operations, and it makes Little's Law a live chart instead of a consultant's slide.

The bigger win is variance. Plants routinely chase the average lead time while the P90 — the worst 10 % of orders, which cause almost all customer escalations — is 3–5× longer than the average. Without per-operation timestamps, that tail is invisible; with them, it becomes the number-one target of the next Kaizen cycle.

FAQ

Is shorter lead time always better?
In most cases yes, but not unconditionally. Cutting lead time by removing safety stock exposes the operation to supplier variability; cutting it by forcing smaller batches without SMED pushes changeover losses up. The honest version of the rule: shorter lead time is better when it is achieved by removing waste (queue, rework, WIP), not when it is achieved by shifting the same waste elsewhere in the value stream. A good indicator is whether the variance shrinks alongside the average — if only the average moves, the improvement is usually cosmetic.

How does lead time relate to OEE?
OEE and lead time are inversely coupled. Every OEE loss — breakdowns, micro-stops, reduced speed, quality rejects — extends the processing window and grows the queue behind the affected process. Plants that improve OEE from 55 % to 75 % typically see lead time drop 30–40 % even without changing the scheduling logic, purely because equipment does what it is supposed to do when it is supposed to do it. This is the most under-appreciated lever in Lean programmes.

What is a good lead-time benchmark for my industry?
Benchmarks mislead more than they help. A tier-one automotive plant with high-volume repetitive parts can realistically run production lead times of hours; a project-based metal-forming job shop is measured in weeks and that is not a bug. The useful benchmark is internal and temporal: the ratio of value-adding time to total lead time (the "process efficiency"), and the trend over time. Best-in-class operations achieve 15–25 % in repetitive discrete manufacturing and 3–8 % in job shops. Anything below 2 % signals structural waste, regardless of industry.

Why does lead time grow non-linearly with utilisation?
This is the queuing-theory answer most plants never hear. In any system with variability, average queue length grows exponentially as utilisation approaches 100 %. Running a work centre at 95 % utilisation does not produce 5 % more queue than 90 % — it produces 30–50 % more queue and therefore a disproportionately longer lead time. The counterintuitive implication: keeping bottleneck equipment at 85 % utilisation often yields shorter and more stable lead times than pushing for 95 %. Toyota understood this in the 1970s; most cost-driven plants still fight it.

How do I report lead time honestly?
Three numbers, always together, never alone: the median (P50), the 90th percentile (P90), and the standard deviation. The median describes the typical order, the P90 describes what your escalation-prone customers actually experience, and the standard deviation describes how predictable the system is. Reporting only the average hides both the outliers and the variance — and those are the numbers that drive customer complaints and expediting costs. Any MES with per-operation timestamps can produce this view in minutes; any plant that cannot produce it is flying blind on the most important operational KPI it has.

Related: Cycle Time · Takt Time · Pull Control · Kanban · Just-in-Time · SMED · TPM · OEE · Value Stream Mapping

About the author

Christian Fieg

Head of Sales at SYMESTIC. Six Sigma Black Belt with 25+ years in automotive manufacturing. Designed and ran JIT/JIS pull loops for Johnson Controls across plants in China, Mexico, the US and Europe, where production lead time was a contract clause, not a KPI. Author of "OEE: One Number, Many Lies". · LinkedIn