A figure without a target is a measurement. A target without a tolerance turns every normal fluctuation into an agenda item, and a meeting that discusses noise soon discusses nothing.
Four numbers make a figure readable, and two more make it honest.
The four numbers
Baseline. The value at the start, with the period it was measured in. Take at least three periods: a baseline from the best month flatters the start and hides the progress that follows.
Target. A value, a date and an owner. Without a date a target is a wish. Targets come from the plant targets and from the cause and effect chain, then get checked against what the process can do. In the example plant the delivery target of 95.0 percent came from the customer scorecards, the lead time target of 8.0 days was the value that delivery target needed, and the first pass yield target of 96.0 percent was what the process had already reached in its three best weeks.
Tolerance. The band that counts as normal, read from the spread of the last twelve periods rather than from taste. Too tight produces weekly drama. Too wide hides a real decline for half a year.
Reaction limit. The value at which an explanation is due. For a figure that should be high it is target minus tolerance. For a figure that should be low it is target plus tolerance. The limit is the strongest rule in the whole system, because it replaces the mood of the meeting with something agreed in advance.
The six headline figures of the example plant
| Key figure | Target | Tolerance | Reaction limit |
|---|---|---|---|
| On-time delivery | 95.0 % | 1.5 points | 93.5 % |
| First pass yield at final test | 96.0 % | 1.0 point | 95.0 % |
| OEE of machining line 3 | 70.0 % | 2.0 points | 68.0 % |
| Conversion cost per unit | EUR 44.00 | EUR 0.80 | EUR 44.80 |
| Manufacturing lead time | 8.0 days | 0.5 days | 8.5 days |
| Skills matrix coverage | 85 % | 3 points | 82 % |
Baselines in the same order: 87.5 percent, 91.2 percent, 62.4 percent, EUR 48.00, 11.5 days and 68 percent.
Two references that keep you honest
The previous year. The same period one year earlier separates a real change from a seasonal one. A plant that improves every spring and falls back every autumn learns nothing from a six-month trend alone.
An external reference, with its source named. There are four reasonable sources, and each has a catch.
A customer requirement tells you the level you must reach to keep the business, but the customer definition may differ from yours. Published industry ranges give orientation, with different scopes behind them, so treat them as a range and never as one number. A sister plant or a neighboring area is usually the most useful comparison, because the definition can be aligned. Your own best three periods show what the current process can do.
A reference tells you the level. Your own trend tells you whether you are getting better. You need both, and the trend first.
When targets change
Targets are reviewed when the product mix, the capacity or the organization changes, not only at the turn of the year. A target that is clearly out of date by March will be quietly ignored by April, and a figure that everybody ignores is worse than no figure.
The opposite failure is just as common: a target reached in month eight that stays unchanged until December. In the example plant the year-end review replaced one target that had been reached with its next stage.
The next part sorts the defined and targeted figures into four views, so that a gain in one place is not quietly paid for in another.