This post has been de-listed
It is no longer included in search results and normal feeds (front page, hot posts, subreddit posts, etc). It remains visible only via the author's post history.
Hi all! Been a full-time IE for a couple of years now, and one of my main responsibilities has been creating labor standards utilizing MOST.
One question that has been burning from the back of my mind ever since the beginning is, is the process I am doing right / necessary?
Right now the expectation as set by management has been 20 hours of observation. No matter how long the job type is, the task can be started and finished within 3 minutes or the task can be 1 hour, the golden rule is get 20 hours of observation and then analyze the data.
The ultimate goal is to be 95% confident w/ an error of or - 5% (pretty standard ask). I’ve done some (albeit) chicken scratch statistics but I always end up with an X thousand number of required observations, which would mean for tasks that have 1 hour or longer it would take thousands of hours of just observations.
Im reaching out and asking for help with this because I believe I can accomplish this observation period with confidence in a drastically lower amount of time than what management has been requesting.
Now I have taken statistics classes and believe I should be capable of answering this question, but anytime I’ve had free time to look into this I’ve always stumbled into analysis paralysis wondering what all the variables are supposed to mean again. This combined with the mental gymnastics from management ensuring 20 hours is the only way just leads me to work on something else and not bother with it now.
Well I decided that this changes today. Wanted to reach out here first, but if there is a better avenue to get this answered then I am all ears. Any help would be greatly appreciated.
Imma leave this comment here as to come back and write a long ass answer. I've had a similar situation like this.
EDIT: MY response has been posted as another comment.
Adding another comment.
First, are you sure you are not falling into a trap?
Scratch that, your management has its head so far up it's ass that it popped outta the neck again. Now that mandatory management derogation is complete, let's head on.
Ok what basis was 20 hours of observation chosen?
For tasks that barely require 10-15 seconds, like tightening a fastener, do you need 20 hours of observation? Pretty sure no. Let's assume the entire task at the work station or the position on a assembly line is tightening multiple fasteners, to torque specs, which can take some time. 10 fasteners, 15 seconds each, is still 150 seconds. That is 480 repetitions in 20 hours, assuming ideal conditions. In real world with shift changes, breaks, you'll be getting around 15-16 hours of actual observations. Now, if management insists on 20 hours of readings, they're even more fucked up in their heads.
Do you need 480 repetitions to set a labor standard for fastener tightening? Probably not. At least for initial. A better approach would be 10 hours of observation with 2.5 hours of maintenance every quarter.
Also, do you set the mean as your labor standard from the observations? Or the median? Or is that set by engineering teams in consultation with supervisors?
For a 150 second job, each observation ranging between 141 - 158, 480 observations.
Standard deviation is 4.685 (I've used Excel rand between(142, 157) for observations).
Sample size, n=480 Sample mean, x=149.22 SD = 4.685 Confidence level = 95%
Confidence intervals are (149.22 /- 0.419), about 0.3%
So, we have now established that 20 hours of observations for a task worth 150 seconds is too much.
List of ALL things you're developing MOST labor standards for, in ascending order of the task time.
Dedicate 20 hours each, but those 20 hours are bankable across tasks.
Distribute the total hours around.
Re-calculate your needs.
Now, in the above example, to reach /-1% interval, you need to drop down to about n=27, so I'm pretty sure you're going wrong somewhere, or missing something.
Similar case randbetween (3600, 5400) Average - 4486.63 SD - 542.1215 CI - 95%
At sample size 35, I'm getting a 4% /- interval.
Tools used: calculator.net/confidence-interval-calculator.html
Excel - randbetween, stddev
Subreddit
Post Details
- Posted
- 5 months ago
- Reddit URL
- View post on reddit.com
- External URL
- reddit.com/r/industriale...
Posted as another comment.