All articles

Lead Time Reduction: Find the Constraint First

By Davina · Published

Key takeaways

  • Lead time is set by the constraint stage, not the average of stage times, so improvements elsewhere move nothing.
  • A one-afternoon walk mapping where work queues, not where it is touched, identifies the constraint.
  • Queue time dominates lead time; touch time at most stages is minutes while work waits days.
  • Match the tactic to the constraint type: capacity, batch size, approval, or information flow.
  • Only fixes at the constraint raise throughput, so scope before quoting and hedge throughput claims until the bottleneck moves.

Lead time reduction works only when you attack the stage that sets the pace: the constraint. Cutting cycle time anywhere else moves nothing, because WIP just piles up in front of the bottleneck. This guide shows how to locate that stage, classify it as capacity, batch size, approval, or information flow, and pick tactics that match it.

What actually sets your lead time?

Lead time is not the average of your stage times. It is set by the constraint: the one stage whose capacity falls below demand. Every other stage has slack, so improving them moves nothing. Until you know which stage governs, any tactic you pick is a coin flip.

Lead time is set by the slowest link, not the average

Little's Law gives you the arithmetic: average WIP divided by average throughput equals average lead time. Because WIP piles up at the constraint, queue time there dominates total lead time, while fast stages contribute minutes. That is why a process with ten stages often has one stage responsible for most of the waiting.

A 20-minute constraint walk: where work queues, not where work happens

Walk the process and record where work sits idle, not where hands touch it. The longest queue marks the constraint. For a typical services workflow, that mapping looks like this: https://improvefirst.com/examples/professional-services-workflow. Name the governing stage before quoting any tactic.

How do you measure lead time without a project?

Timestamp handoffs. For one week, note when each stage starts and finishes, including the waits between them. No simulation, no process-mapping workshop. The timestamps alone tell you which stage holds the work longest, and that stage is almost always your constraint.

Queue time vs touch time: the 90/10 split

At most stages, touch time is minutes and queue time is days. Measure both separately: touch time is when someone is actively working on the item, queue time is everything else. In practice, queue time dominates total lead time, so cutting touch time at a non-constraint stage changes almost nothing. The payoff comes from the stage where work waits longest, which is why we say test the bottleneck payoff before simulation (https://improvefirst.com/learn/bottleneck-payoff-before-simulation).

Sizing the demand gap at the suspected constraint

Compare the constraint's available capacity with the demand arriving at it, in the client's own units: orders per day, cases per week, approvals per shift. The gap between demand and system capacity tells you whether the constraint is a capacity problem, a batch-size problem, or an approval problem, and that determines which tactic applies in the next section.

Which lead time reduction tactic works on which constraint?

A tactic only cuts lead time if it attacks the stage that governs the system. Goldratt's core claim in The Goal is that total throughput is set by the constraint, so effort anywhere else changes little. Match tactic to constraint type before promising a client anything.

Tactics matched to constraint type

TacticFixes this constraintDoes nothing for
WIP limitsQueue-driven lead time at a capacity constraintApproval or information constraints
Batch reductionBatch-size constraintsA stage with no queue
Offloading the constraintCapacity constraintsNon-constraint stages
Scheduling / takt alignmentDemand gap or sequencing constraintsBatch-size constraints
Approval streamliningApproval constraintsCapacity constraints

The support ticket capacity example shows the pattern: a queue at one stage dominates lead time, and WIP limits there beat any change elsewhere.

Why offloading the constraint beats speeding up everything else

Speeding a non-constraint stage adds no throughput; the backlog just moves to the constraint. Offloading or adding capacity at the constraint is the only move that raises system capacity. Run the constraint check before quoting, or you risk charging for work with no measurable effect.

What does a worked constraint calculation look like?

The non-constraint trap: a 50% cut that saves zero days

Take a five-stage process with cycle times of 4, 10, 3, 5, and 2 days, sequential, no overlap. Stage 2 is the constraint: system capacity is one unit every 10 days, and lead time is 24 days. Cut Stage 1 from 4 days to 2, a 50% improvement, and lead time stays at 24 days. The queue in front of Stage 2 simply grows, WIP rises, and the client sees no delivery benefit at all. Now cut Stage 2 by 20%, from 10 days to 8, and lead time drops by the full 2 days, because the constraint governs the whole system. That asymmetry is the entire argument for finding the bottleneck before promising anything.

Testing the fix before promising it to the client

Run both scenarios before you quote. A tool like the Constraint Impact Calculator (https://improvefirst.com/learn/constraint-impact-calculator) lets you model the cut, confirm the lead time change, and predict where the constraint moves next, likely to Stage 4 at 5 days. If Stage 2's fix only works with overtime you have not costed, you want to know that before the client does.

How does lead time reduction differ in manufacturing vs services vs inventory?

The constraint changes shape by context. Manufacturing hides it in machine capacity and changeovers, services hide it in approvals and handoffs, and inventory hides it in order quantities and supplier variability. Diagnosing the wrong type wastes the client's money and your credibility.

Lead time reduction in manufacturing: changeovers and takt

The constraint is usually a machine whose capacity is eaten by setup time, not run speed. Shingo's SMED approach attacks changeovers directly, separating internal from external setup so the bottleneck machine spends more time producing (Shingo Institute on SMED). Check whether WIP in front of the constraint masks a takt mismatch upstream.

Lead time reduction in inventory management and the supply chain

Here lead time is set by order quantity rules and supplier variability, not by any internal stage. Large batches look efficient on unit cost but inflate average lead time and WIP downstream. The work is quantifying demand gap against replenishment frequency before touching supplier terms.

Service processes: the approval bottleneck

Service constraints are rarely capacity-limited; they are information-limited. Work queues behind a single approver or an incomplete handoff, and no amount of staffing fixes a missing decision right. Map who can say no, then measure how long they take.

The Constraint Impact Calculator scopes all three contexts, but it assumes you can express the constraint as a measurable stage. Purely behavioural service constraints, where the "stage" is a person's judgement, still need observation before any calculator helps. See https://improvefirst.com/legal/disclaimer for the limits of modelled estimates.

How do you scope a lead time reduction engagement before quoting?

The scoping pass: one afternoon, one constraint hypothesis

Spend half a day mapping the process, measuring cycle time and WIP at each stage, and naming the one stage that sets lead time. Write it as a hypothesis: "the constraint is the approval step, not the assembly line." Test it against the client's demand gap: if the constraint stage's capacity covers demand, look upstream for batch size or information flow instead. A hypothesis you can falsify in an afternoon is worth more than a two-week discovery phase the client has to pay for before seeing value. Then put the pass in writing. A client-ready report should state the constraint hypothesis, the measurements behind it, the one tactic matched to that constraint type, and the estimated throughput gain if the constraint moves. That document is your quote's justification: the deeper engagement becomes "verify and exploit the constraint," not "lean project, six months."

When to walk away: constraints that are policy, not capacity

If the constraint is a rule the client's leadership won't revisit, like a mandatory dual sign-off or a fixed batch quantity set by finance, no amount of your work on the floor changes lead time. Scope these out loud: "the constraint is policy, and fixing it is a management decision, not a consulting engagement." Walking away from that quote protects your margin and your reputation.

Which tools help, and where does each fall short?

Comparing the options on the same criteria

For scoping work, judge tools on four criteria: can you model the constraint quickly, can you show the client where throughput is lost, can you test one change cheaply, and can you hand over a client-ready summary. A spreadsheet capacity model wins on cost and flexibility but hides dependency logic in formulas only you can read. ProcessModel supports fuller simulation, which matters when you need to show a constraint moving after you exploit it.

Where the Constraint Impact Calculator falls short

Improve First's calculator is built for the first two criteria: find the constraint, test whether fixing it lifts system output. It stops short on dynamic behaviour. It will not simulate arrival variability, queueing, or supplier lead time variability, which can push inventory lead time well beyond the quoted average (safety stock and lead time variability). When the client's constraint shifts with demand mix or batch policies, use a simulation tool instead, and quote accordingly.

What benefits can you credibly promise a client?

Benefits that follow from the constraint math

Only improvements at the constraint raise throughput. Offloading a non-constraint stage changes nothing for the client's output, which is why you should scope before quoting: a Constraint Impact Calculator run on their process shows whether the constraint is capacity, batch size, approval, or information flow, and therefore which benefit is even available.

WIP and backlog respond more predictably than throughput. Cutting batch size at the constraint shortens lead time and drains WIP even if total output holds steady, so those are defensible early wins while the demand gap closes over longer horizons.

What to promise and what to hedge

Promise what the math supports: a bounded lead time reduction at the constraint, expressed in days or hours, and a WIP drop you can measure within one cycle. Hedge throughput claims as a hypothesis until the constraint has moved, because once you fix the current bottleneck, the next one sets the ceiling and your original projection stops applying. State that re-scoping after the first improvement is part of the engagement, not an admission the first estimate was wrong.

FAQ

How long does lead time reduction take to show results?

If the fix targets the constraint, a measurable lead time drop appears within one cycle, and a WIP reduction is often visible even sooner. Fixes at non-constraint stages show nothing at any horizon, because the queue just shifts. That asymmetry is why the one-afternoon scoping pass exists: it tells you which change will pay back in days rather than quarters.

Can you reduce lead time without increasing capacity?

Yes, in three cases. Cutting batch size at the constraint drains WIP and shortens lead time even when output holds steady. Streamlining an approval constraint removes waiting without adding staff. And WIP limits shorten lead time by shrinking the queue, though throughput stays capped until the constraint itself gets more capacity. What never works is speeding up a stage that already has slack.

What is the difference between lead time and cycle time in this context?

Cycle time is how long one stage actively takes to process an item; lead time is the total from request to delivery, including every wait between stages. Because queue time dominates lead time, the two diverge sharply: a stage with a 10-day cycle time can hold work for weeks once the backlog in front of it is counted.

Does reducing batch size always cut lead time?

No. Batch reduction works when large batches are inflating WIP at or upstream of the constraint. If the constraint is a capacity shortage, an approval bottleneck, or a policy like a finance-mandated order quantity, smaller batches just add changeovers or handoffs without moving the queue. Check where work actually waits before applying the tactic.

How do I calculate the reduction of lead time percentage for a client report?

Measure the baseline lead time from timestamps before any change, then measure again after the constraint fix, using the same start and end points. Subtract, divide by the baseline, and multiply by 100. Only quote the figure if the change touched the constraint; a percentage cut at a non-constraint stage is real arithmetic but zero client benefit.