6.1 Process Flow Metrics and the Hidden Factory

The core flow metrics

Takt versus cycle time

$\text{Takt time} = \frac{\text{Available production time}}{\text{Customer demand}}$

If a line runs two 8-hour shifts with 45 minutes of breaks per shift and demand is 870 units per day:

$\text{Takt} = \frac{2 \times (480 - 45) \text{ min}}{870} = \frac{870}{870} = 1.0 \text{ min/unit}$

Takt is a requirement, not a measurement. Cycle time is a measurement of what the process actually does. Comparing the two at each step is the standard line-balancing analysis:

• Step cycle time greater than takt: the step cannot keep up; it constrains the line.
• Step cycle time much less than takt: the step is under-utilized; consider combining stations.
• All steps close to but under takt: balanced line.

The bottleneck is the step with the longest cycle time. Throughput of the whole process equals the throughput of the bottleneck, which is why improving a non-bottleneck step changes nothing.

Little's Law

$WIP = \text{Throughput} \times \text{Lead time} \quad \Longleftrightarrow \quad \text{Lead time} = \frac{WIP}{\text{Throughput}}$

Little's Law is the most useful relationship in the Measure phase because it holds for any stable process regardless of its internal structure. Two consequences:

1. Lead time can be reduced by reducing WIP alone, with no change to the process itself. Halving the number of orders released into a stable process halves the lead time.
2. WIP is a controllable input. Releasing more work into a system already at capacity increases WIP and therefore lead time, without increasing throughput at all.

Worked example. A claims process holds 480 open claims and completes 60 per day.

$\text{Lead time} = \frac{480}{60} = 8 \text{ days}$

If a WIP cap reduces open claims to 240 while throughput stays at 60 per day, lead time falls to 4 days. Nothing about the work changed; only the queue did.

Touch time and process cycle efficiency

$\text{Process cycle efficiency} = \frac{\text{Touch time (value-added time)}}{\text{Total lead time}} \times 100$

A loan application with 42 minutes of touch time and a 6-day (2,880 working-minute) lead time has an efficiency of 1.5%. Transactional processes commonly fall between 1% and 10%; a well-run lean manufacturing cell may reach 25% or more. The number is worth computing early because it tells the team where the opportunity is: at 1.5%, working on the 42 minutes of touch time is pointless, and all the value is in the 2,838 minutes of waiting.

Identifying the constraint

The Body of Knowledge specifically asks that these metrics be used to determine constraints. A practical sequence:

1. Compute takt time from actual customer demand.
2. Measure cycle time at every step, with enough observations to see the variation, not just the mean.
3. Plot cycle time by step against the takt line.
4. Look for where WIP and WIQ accumulate -- inventory piles up in front of the constraint.
5. Confirm with utilization: the constraint runs near 100% while others idle.

The physical evidence usually settles it before the arithmetic does. Queues form immediately upstream of the bottleneck and starvation appears immediately downstream.

The hidden factory

A hidden factory is the set of undocumented activities -- rework, retest, workarounds, informal inspection, chasing missing information -- that a process performs routinely but that appears in no routing, no standard time, and no yield report.

Why it matters to each metric:

The classic symptom is a process that reports 99% yield but requires 15% overtime to meet a schedule it was sized for. That gap is the hidden factory.

Finding it

• Walk the process and count every unit that moves backwards. Rework almost always travels against the primary flow.
• Compare standard routing time with actual labour hours per unit. The difference is unrecorded work.
• Ask operators what they do when something is wrong, not whether defects occur. The answer is usually a well-established informal procedure.
• Measure first-pass yield at each step, defined as units passing with no rework of any kind, rather than final yield after correction.
• Look for uncounted inventory: the reject bin that gets reworked at shift end, the "pending clarification" folder, the parts on the maintenance bench.

The measurement fix is to define first-pass yield as right first time with no touch-up, and to make rework visible by requiring it to be recorded with a reason code before the unit re-enters the flow. Making rework visible almost always reduces reported yield in the first month, which must be explained to the sponsor in advance or it will be read as the project making things worse.