Most plans are late on the day they are released. Not because people work slowly, but because the plan assumes hours that do not exist.
Capacity planning compares two numbers per work center and week: what the work center can really do, and what the plan asks of it.
Available hours
Start with the nominal hours and subtract what every week takes away. For five CNC machining centers on two shifts:
- nominal: 5 machines x 2 shifts x 5 days x 8 hours = 400 hours,
- breaks with machine stop: 25 hours,
- planned maintenance and cleaning: 15 hours,
- meetings and start-up: 10 hours.
Available: 350 hours per week. Use averages of at least thirteen weeks, not the best week of the year. Planning against 400 hours would make every week look feasible and every week end late.
Load
Load is the work the plan puts on a work center, in hours:
quantity x standard minutes, plus number of setups x setup minutes, divided by 60.
In week 1 the plant needs 500 small, 260 medium and 110 large valve blocks. At CNC machining that is 500 x 14 + 260 x 22 + 110 x 38 = 16,900 minutes, or 281.7 hours. Fifteen setups of 45 minutes add 11.25 hours. Total: 292.9 hours against 350 available, a utilization of 84 percent.
Read the table
Do the same for every work center and the next six weeks:
| Work center | Available h | Load week 1 | Utilization |
|---|---|---|---|
| Saw | 70 | 33.2 | 47 % |
| CNC machining | 350 | 292.9 | 84 % |
| Washing and deburring | 75 | 38.8 | 52 % |
| Assembly | 300 | 228.4 | 76 % |
| Test | 150 | 102.7 | 68 % |
The table answers two questions at once. Which work center limits the plant? Here CNC machining. And which weeks will not work? In week 5 the order book needs 357 CNC hours, 102 percent.
Three ways to close a gap
When load exceeds capacity, there are three levers, in this order:
- Move volume: pull orders into a week with free capacity.
- Change capacity: a Saturday shift, a third shift on two machines, temporary staff.
- Change demand: agree a new date before it is promised, or outsource simple parts.
Each lever has a lead time. Moving volume takes days, adding a shift takes one to four weeks. That is why the load view must look six weeks ahead, not one.
Master data decides
The load is only as good as the standard and setup minutes behind it. Compare confirmed actual times with standards every quarter and correct every operation that is off by more than 10 percent. A precise calculation on wrong data produces wrong dates with great confidence.