Customer orders do not arrive evenly. One week brings 850 valve blocks, the next 1,060. If the plant produces exactly what was ordered for each week, the shop floor swings between idle machines and Saturday shifts.
Leveling takes the swings out before they reach the floor.
Level the volume first
In the example plant, CNC machining is the bottleneck. Its utilization over six weeks:
| Week | Blocks ordered | CNC utilization |
|---|---|---|
| 1 | 870 | 84 % |
| 2 | 900 | 85 % |
| 3 | 990 | 96 % |
| 4 | 850 | 83 % |
| 5 | 1,060 | 102 % |
| 6 | 920 | 87 % |
Week 5 cannot be done in normal hours. Week 4 has room. Pulling 40 small and 20 large blocks forward from week 5 into week 4 changes the picture: week 4 rises to 89 percent, week 5 falls to 96 percent.
The customer receives these blocks a few days early, or the plant holds them in finished goods for a few days. Both cost less than overtime, expediting and the orders pushed back by a peak.
Why 85 to 95 percent
A bottleneck at 100 percent has no room for anything: a breakdown, a quality hold, a missing program. Every disruption becomes a late order. A bottleneck at 70 percent wastes the most expensive capacity in the plant.
Planning the bottleneck between 85 and 95 percent leaves a margin for the disruptions every plant has. In the example, the bottleneck loses about 8 percent of its time to disruptions. A plan at 100 percent is late before the week starts.
Then level the mix
Volume leveling decides how much per week. Mix leveling decides how the families are spread inside the week. Instead of all large blocks on Monday and all small ones on Friday, each day carries a similar share of each family.
Mix leveling needs smaller lots and more setups. That is why it comes second, after setup times at the bottleneck have been reduced. A useful sequence of maturity:
- weekly volume against bottleneck capacity,
- weekly mix of families with lots of at most one week,
- a fixed daily mix pattern once setups at the bottleneck are below 20 minutes,
- a pitch of a few hours with Kanban loops for stable products.
Who decides
Leveling is a planning decision with a customer effect. The planner proposes the moves in the weekly planning review, the production manager agrees, and sales confirms where a customer date changes. Record the moves, so that the next peak is not solved by the same discussion again.
The result is less visible than a new machine. It is often worth more: fewer overtime shifts, fewer rush orders and a floor that runs at a pace people can keep.