Ask a production manager where the time in the routing came from and you usually get a pause. Somebody measured it, years ago. Somebody else copied it when the variant was created. It has been in the system ever since, and four separate decisions read it every week.
The four readers of one number
A standard time is used in the quotation, in the product cost, in the capacity plan and in the takt of a line. None of these four asks how the number was produced. All four behave as if it were exact.
That is why a wrong time does not announce itself. It appears as an investment request, a lost order, a station that always looks slow, or a takt the line cannot hold. The cause sits in a field nobody has opened for years.
What a standard time actually contains
Five parts, and leaving one out makes the number incomparable with every other time in the plant.
The observed time is what the watch showed for a defined operation. The rating compares the pace seen with a written reference pace, and turns the observed time into a basic time. The allowances cover recovery and short unavoidable interruptions, applied to manual work. The machine time is read from the equipment, never estimated. The setup share is the setup time per lot divided by the planning lot size.
Three words carry the weight: defined, reference pace, allowances. Change the method and the time changes. Leave out the allowances and the time can never be held for a whole shift.
Six times that summed to the right total
Area M2 assembles and tests hydraulic control blocks on six stations in line, two shifts, 230 working days. The six station times in the system dated from 2019 and summed to 14.10 minutes per unit. After the stations were measured, the new times summed to 14.10 minutes as well.
| Station | In the system | Measured | Deviation |
|---|---|---|---|
| S1 Cartridge pre-assembly | 2.10 | 2.34 | +11.4 % |
| S2 Housing preparation | 1.80 | 1.92 | +6.7 % |
| S3 Valve block assembly | 3.60 | 2.83 | -21.4 % |
| S4 Cover, torque and marking | 2.40 | 2.58 | +7.5 % |
| S5 Leak and function test | 3.00 | 3.36 | +12.0 % |
| S6 Pack and label | 1.20 | 1.07 | -10.8 % |
The total was right because two errors offset each other. Nobody plans with a total.
What the errors cost
S3 at 3.60 minutes looked like the slowest station, so it was managed, and an investment request for a second assembly station worth EUR 95,000 was on the table. The real constraint was the test bench S5 at 3.36 minutes, which nobody watched. The published takt of 3.18 minutes sat below the bottleneck time, so a queue grew in front of the test bench by about twelve blocks a day. A variant with a longer test cycle was costed with 3.00 minutes instead of 4.11, which is EUR 1.07 per unit and EUR 6,420 a year on 6,000 units.
And for two years the test bench reached 70.5 percent of its standard hours while the assembly station reached 103.1 percent. One team was asked about performance. Both numbers were artifacts.
What a standard time is not
It is not a target for a person: it is the time a trained person needs with the method as it is today. It is not the lead time, because waiting, transport and queue time sit outside it. It is not the machine running time, and where a machine runs without the operator the two have to be separated. And it is not permanent. A standard time belongs to a method version and expires when the method changes or the review date passes.
Write that down on one page before the first study. Every time that follows will be comparable with every other.