English is currently available.
Industrial Engineering DISCUSSION
I timed an operator doing a manual packing task with a stopwatch: ten cycles with a mean of 0.80 minutes and a standard deviation of 0.08 minutes. I judged the operator to be working a bit faster than normal pace, around 110 percent, and the company uses a 15 percent allowance.
How do I turn this into a standard time and an expected output per shift, and were ten cycles enough to trust the result?
Three steps. Observed time is the mean of the timed cycles, 0.80 min. Normal time adjusts it to the pace of a qualified operator working at normal performance: normal time = observed time × rating = 0.80 × 1.10 = 0.88 min. Standard time adds allowances for personal needs, fatigue and unavoidable delays: standard time = 0.88 × (1 + 0.15) = 1.012 min per unit.
Output at standard over 480 minutes of working time is 480 / 1.012 = 474 units. If the allowance already includes rest breaks, do not subtract the same breaks from the shift a second time.
Check how your company defines the allowance, because two conventions exist. If the 15 percent is a percentage of normal time, the formula above applies. If it is a percentage of the working day, the formula is standard time = normal time / (1 - 0.15) = 0.88 / 0.85 = 1.035 min, and the shift output is 464 units. The difference is 2 percent, which matters when the standard is used for pay or capacity planning.
Whichever is used, write it on the study sheet along with the method, layout, tools and materials, since the standard is only valid for that method.
For the number of cycles, the usual formula is n = (z × s / (k × m))², where s is the sample standard deviation, m the sample mean, k the acceptable error as a fraction of the mean and z the normal value for the confidence level. For ±5 percent at 95 percent confidence: n = (1.96 × 0.08 / (0.05 × 0.80))² = (3.92)² = 15.4, so 16 cycles. Your ten cycles are not quite enough for that precision; they give about ±6 percent.
With so few readings it is more correct to use the t value instead of z, which pushes the requirement up a little further.
Rating is the weakest part of a stopwatch study because it is a judgement. Practical ways to make it more defensible: time a qualified, trained operator who knows the method, not the fastest or the newest; break the job into short elements and rate each element; remove abnormal readings only when the cause is identified and recorded, such as a dropped part; and have observers practise rating against benchmark films of known pace.
Tell the operator what is being done and why. A study carried out covertly produces a distorted pace and a standard nobody accepts.
Two alternatives avoid rating altogether. Predetermined motion time systems such as MTM build the normal time from tabulated times for basic motions like reach, grasp and position, in units of TMU, where 1 TMU is 0.00001 hour or 0.036 second. They can set a standard before the job exists, but need a trained analyst.
Work sampling estimates how time is divided between activities from random observations. The number needed is n = z² × p × (1 - p) / e²; to estimate an activity that takes about 20 percent of the time to within ±3 percentage points at 95 percent confidence, that is 3.84 × 0.16 / 0.0009 = about 680 observations.
Back to Industrial Engineering community
