A symptom does not identify its mechanism.
The queue can grow because labor is insufficient. It can also grow because available labor is waiting, switching, reworking, or processing the wrong work.
Staffing is visible, measurable, and purchasable, so it often becomes the default explanation. But adding people upstream of a technical bottleneck, into an unstable process, or without the required skill mix can increase work in process without increasing completed output.
A credible hypothesis states what evidence should appear if it is true—and what result would falsify it.
Six mechanisms deserve a test before the hiring request.
True labor capacity
Signal: productive stations remain continuously loaded, service time is stable, and added trained hours create proportional completed output.
Skill mix
Signal: work waits at specific certifications or decision rights while other labor has availability.
Process flow
Signal: queues concentrate at handoffs, batching points, approvals, or exception loops rather than across the entire process.
Quality loss
Signal: rework, reopen rates, defects, or first-pass yield explain a material share of consumed capacity.
Schedule alignment
Signal: capacity and demand exist at different hours, days, sites, or service windows.
Technology or policy
Signal: downtime, latency, access, rules, or approval thresholds cap output while labor waits or works around the system.
Move from suspicion to a fundable decision.
- Map the work at queue level.Measure arrivals, completions, aging, WIP, service time, and waits by step—not only at the final output.
- Rank competing mechanisms.Use observed queue location, utilization, yield, downtime, and schedule patterns to prioritize tests.
- Write a falsifier.State what evidence would show the leading explanation is wrong before collecting more confirmatory data.
- Run the smallest discriminating test.A controlled shift, dedicated skill cell, batch-size change, downtime window, or quality intervention should distinguish mechanisms.
- Measure completed output.Do not declare success from activity, staffed hours, or local utilization if the end-to-end queue does not move.
- Scale only after the response is repeatable.Convert the test into a staffing, process, technology, or governance decision with an accountable owner.