Calculating and Testing a WIP Target

Executive summary. Excess work in process (WIP) ties up capital and can extend lead times; too little inventory in the wrong place can leave downstream operations starved when conditions vary. Little's law relates average WIP to throughput and average flow time. It does not, by itself, determine the minimum inventory or a safety buffer. This article presents the calculation, a complete numerical example, and a way to test a proposed operating target using plant data.

---

1. The Business Problem: WIP That Has Never Been Calculated

Every part waiting between stages represents material, work already performed, and cash not yet recovered. But cutting WIP indiscriminately can create other problems: after a short stoppage, an out-of-specification part, or a defective batch, the next operation may run out of work.

A buffer set by habit—“two pallets just in case”—does not explain the relationship between inventory and flow. Equally, a calculated average is not automatically a minimum that protects every stage.

The management question is specific: how much WIP is needed, and where, to support the required output with an acceptable level of starvation, blocking, and lead time? A measured and tested target can put working capital, space, and delivery discussions on a clearer footing.

2. The Base Relationship: Little's Law

Little's law (Little, 1961, “A Proof for the Queuing Formula: L = λW,” Operations Research) relates the average number of items in a system to its effective throughput and the average time an item spends inside it:

WIP = throughput × flow time

• Throughput: completed units per hour, measured on a consistent system boundary.
• Flow time: average hours from entry into that boundary to exit, including processing and waiting.
• WIP: average units present within that same boundary.

A quick example: if throughput is 12 units per hour and average time in the system is 5 hours, average WIP is:

12 × 5 = 60 units

This describes compatible averages. It does not establish that 60 is the lowest feasible WIP: if waiting time can be reduced, the same throughput could be associated with a shorter flow time and less WIP.

Two distinctions matter. First, bottleneck capacity constrains sustainable throughput, but a station rated at a given speed does not prove that the line actually achieves that output. Second, Little's law is not restricted to a variability-free process. Waiting caused by variability is already included when measured flow time includes queues. Consistent boundaries, time bases, and an appropriate long-run flow balance are essential.

3. A Variability Margin: A Candidate Buffer to Test

Real plants have micro-stoppages, changeovers, quality issues, and absences. These can affect starvation, blocking, throughput, and flow time. Little's law alone does not prescribe the inventory buffer that will achieve a specified service or production objective.

The source calculation can be used as a planning scenario:

Proposed operating WIP target = (throughput × flow time) × (1 + margin)

This retains the multiplication but should not be called a proven minimum. If the flow time already includes existing queues and delays, adding a percentage can double count their effect rather than identify necessary protection.

The percentages in this example are illustrative: 10% for a more stable scenario, 25–30% for a more variable scenario, and 20% for the worked calculation. They are not established industry recommendations. Select or revise a target using the process data, buffer locations, and performance objective.

With the preceding 60-unit baseline:

60 × 1.2 = 72 units

That is a candidate inventory target, not a mathematical guarantee that the line will avoid stoppages. Increasing inventory can also change flow time, so the three quantities must be checked again after implementation.

More buffer than needed ties up capital; too little or poorly located protection can increase starvation. Measure the trade-off instead of choosing a percentage by preference.

4. Validating the Calculation with Plant Data

A calculation based on estimates is a hypothesis. Examine three areas before turning it into an operating policy:

1. Actual throughput by stage. Measure completed output and distinguish achieved rates from nominal machine capacity. Include a representative period and the relevant product mix. Investigate the constraint rather than assuming the fastest station determines the system's output.
2. Where work accumulates and where resources stop. Work sampling can classify observed activity states—for example VA/NVA/IMP, meaning value-added, non-value-added, and nonproductive/delay under a defined taxonomy. These proportions help identify material-starved or blocked resources. However, observing workers does not directly count WIP or measure a part's entry-to-exit flow time. Collect item counts and timestamps, or appropriate system records, alongside sampling.
3. The required observation count. Select precision and confidence for each relevant proportion. WorkSamp includes a sample-count calculation for a 95% confidence setting; a confidence level alone does not determine the number required. Coverage, dependence, and classification quality still matter.

When the detailed time of a particular operation is needed to assess capacity, stopwatch time study can complement this work. See Work Sampling vs. Time Study.

After selecting a candidate WIP limit, monitor actual throughput, average WIP, flow time, starvation, blocking, and quality. Adjust the total and its distribution between stages rather than treating the first result as final.

Simulated Practical Case

This example was constructed for teaching purposes. It does not describe an actual WorkSamp implementation or a particular customer.

A two-stage assembly cell is expected to deliver 10 units per hour. The first stage, cutting, has a rate of 14 units per hour, while assembly reaches only 10. Assembly therefore constrains the stated capacity. Using 10 units per hour as the calculation throughput also requires checking that the whole cell actually sustains that output.

Average end-to-end flow time, from entry into cutting to exit from assembly, is 4.5 hours.

• Average WIP implied by Little's law: 10 units/hour × 4.5 hours = 45 units.
• Candidate target with an illustrative 20% margin for the stated stoppage and changeover scenario: 45 × 1.2 = 54 units.

A work sampling study also reports material waiting at assembly in 18% of observations. This is an illustrative activity proportion, not evidence that the correct buffer can be calculated directly from that percentage.

The proposed trial compares a suitably distributed target of 54 units with the previous 85 units. It tests whether inventory can be reduced while sustaining output and controlling starvation. Neither the 18% observation result nor the formula proves that waiting will disappear. Any claimed capital release requires the actual inventory change and valuation; none is asserted as a measured customer outcome here.

Checklist for Your Line's WIP Target

• Is reference throughput measured, with nominal bottleneck capacity treated as a constraint rather than an achieved result?
• Does flow time cover the same boundary and include every relevant stage and queue?
• Is the variability margin an explicitly stated hypothesis supported by process evidence?
• Does buffer placement address where starvation and blocking occur?
• Is there a periodic review when product mix, volume, or operating conditions change?
• Are WIP counts and item flow times measured separately from sampled worker activity?

Conclusion

A useful WIP target is a business decision grounded in measurements. Throughput multiplied by flow time establishes a relationship between averages; it does not on its own calculate a minimum or prescribe an extra percentage.

Use that relationship to check consistency, then test buffer size and location against the required output, lead time, and operating risks. This is how a spreadsheet scenario becomes a defensible plant decision.

To examine where resources wait and which conditions need further measurement, request a demonstration at WorkSamp. If you are deciding which measurement method to use, start with Work Sampling vs. Time Study.

Primary reference

John D. C. Little (1961), “A Proof for the Queuing Formula: L = λW,” Operations Research, 9(3), 383–387. The reference establishes the relationship between averages; the illustrative margins and operating targets in this article are not derived from that theorem.