A team's busiest hour is not its dependable capacity. People take breaks, systems stop, jobs differ and some work must be repeated. A useful estimate states which of those factors are included and which remain uncertain.
Define the unit before counting
For a fictional repair service, the unit might be a correctly repaired device rather than an opened ticket. If diagnosis, repair and inspection all process those devices, their rates can be compared. If one step counts individual components and another counts whole devices, the numbers first need a conversion.
Choose a period that matches the decision. An hourly rate helps with a same-day queue. Weekly capacity may help plan bookings, provided the mix of jobs and staffing are comparable. Do not compare a technician's rate per working hour with demand per calendar day without adjusting the periods.
A worked estimate
Suppose one repair bench is staffed for eight hours a day. Breaks, setup and routine maintenance leave six hours available. Within those hours, the team can attempt ten standard repairs per hour, and nine out of ten pass final inspection without further work.
| Component of the estimate | Fictional value |
|---|---|
| Staffed time | 8 hours/day |
| Available processing time | 6 hours/day |
| Attempts per available hour | 10 |
| First-pass acceptable share | 90% |
| First-pass good output | 6 × 10 × 0.90 = 54/day |
The 54 is a first-pass output estimate. It is not a complete capacity model if failed devices return to the same bench and consume additional time. Either include that rework in an observed effective rate, or model its time separately. Do not apply the quality adjustment twice to a rate already measured in good output.
Compare the entire path
Imagine diagnosis can complete 70 devices per day, the repair bench 54 and inspection 65. With enough matching demand, the repair bench is the smallest stated rate. Increasing diagnosis to 90 does not raise the simplified system ceiling above 54.
This comparison assumes a serial process, a consistent job mix and no hidden shared resource. If the same person runs diagnosis and inspection, their separate maximum rates cannot necessarily happen together. Check the staffing calendar before treating the stations as independent.
Use the bottleneck tool to make a first comparison. Treat its result as a hypothesis to verify against actual completed output, not a promise of what will happen tomorrow.
Collect evidence over a representative period
Record available time, good completions, rework and the mix of jobs. Include both ordinary days and the disruptions that matter to the decision. Compare the estimates with the work actually delivered. When output is lower, ask whether demand, materials, staffing or a mistaken rate explains the difference.
Avoid using a single quiet day to declare spare capacity or one exceptional day to promise a permanent rate. A range can be more honest than a precise-looking average. Keep a note of the conditions under which each estimate was observed.
Your next small test
Pick one stage and write its unit, available time and effective rate. Identify one recurring loss of useful time. Change that loss for a short observation period and measure good completed output across the whole service. For a stronger diagnosis, first map the workflow and examine why the queue grows.
Sources and limits
The focus on the limiting step draws on the Theory of Constraints Institute's Five Focusing Steps. The repair business, rates and calculations are original fictional HBW examples. The method does not estimate profit or describe every variable, parallel or shared-resource system.