- June 4, 2018 at 11:08 pm #56011
A project gave us the following outcome (unit=minutes):
Mean: 15, Median: 11, Std Dev: 12
Mean: 10, Median: 10, Std Dev: 7
When we have a normal distribution, we take the difference in mean, annualize it, and multiply it with the headcount to get the annual person hours savings. What would you suggest in this case considering it is non-normal and there is a substantial change in std dev?June 5, 2018 at 2:27 pm #202629
The average works since $ are related to actual $ spent which is number of folks….mean works mathematically better and try and get any finance guy to use a median.
Think of this. If the average weight of gold put on a circuit board was X grams and it was reduced by 10% or to a total of 0.9*X, then the savings is 0.1*X*($/unit of mass)*# of units–independent of the median even if was a non-normal distribution.June 11, 2018 at 5:25 am #202640
If due to higher Standard deviation Spread goes outside USL and LSL you can use reduced standard deviation as reduction of number of defect and can be converted into savingJune 13, 2018 at 8:44 am #202657
@mrinal222 Chris Seider gave you a good suggestion if you want to do labor hours. The problem with labor hours as a general measure doesn’t mean there was an actual dollar savings. If I have 2 people doing a job and after I do some improvements I need 1.5 people then there is no dollar savings (particularly in the eyes of an accountant) because I still need 2 people. This leads to the discussion of FTE’s (Full Time Employees).
Just my opinionJune 13, 2018 at 9:12 am #202658
@cseider Appreciate your prompt response. I think I get your answer only partially. In the gold example, when you consider a sample to get the mean weight, suppose you have one data point with an extreme value. That data point is very likely to skew the mean. The mean will not remain representative anymore, right? Does it still remain valid? It will if we do not consider the outlier points though.June 13, 2018 at 9:25 am #202660
@mike-carnell Thanks for your response. I have a different take on the subject. Theoretically, FTEs could be broken down into many units. If I save 0.5FTEs, I can get him/her to do a job for which the organisation may have had to hire an additional resource. That’s cost avoidance. If I had a job that required half a resource’s time and the other staff members didn’t have any bandwidth,I’d have to hire a new resource.The savings of.5 FTE can help here.June 13, 2018 at 11:44 am #202664
@mrinal222 You may have a different take on FTE’s but you are going to have a difficult time selling it to anyone from accounting that needs to sign off on what you claim for benefits. “Soft Benefits” are like a footnote nobody reads.
I will guarantee you that if for any reason you are audited by a major accounting firm it won’t fly as a benefit.June 13, 2018 at 1:05 pm #202667
@mike-carnell brings good point on FTE’s.
One must always use the mean–remember your savings per (part) is the total change in the input (cost) divided by the total pieces/processes made times the cost/input.
I used to think hard about using medians for savings for non-normally distributed problems solved but the math doesn’t work out AND try getting a finance guy to use medians in their benefit calcs.
You must be logged in to reply to this topic.