|
iomnet
|
|
|
System Administrator
Total Posts: 87
Joined: 12 Feb 07
Member #1
|
I wish to know as to whether there is a standard definition of how accurate stock should be in order to run an automated replenishment system? If so, how is this measure calculated, as I have read of a number of different ways that it can be calculated?
|
|
iomnet
|
|
|
System Administrator
Total Posts: 87
Joined: 12 Feb 07
Member #1
|
The optimum way is by lagrangian multipliers
John C
|
|
iomnet
|
|
|
System Administrator
Total Posts: 87
Joined: 12 Feb 07
Member #1
|
Any stock system must have a feedback and correction control loop. Measure the ratio of error correction messages to good matching measurements; e.g. in a cycle counting system, the number of stock adjustments to matching counts.
More simply, I can count the number of stock adjustment transactions every month and compare them to the number of warehouse items or the number of stock movements.
Any target percentage is arbitrary. But we know that any system with below 96% accuracy is doomed. As Dr. Deming says, arbitrary numerical targets are nonsense. The point is to maintain your statistical measures with a view to removing problems and improving performance. You can argue about the law of diminishing returns when you get to 99.9%
Or wear a pedometer.
Henry B
|
|
iomnet
|
|
|
System Administrator
Total Posts: 87
Joined: 12 Feb 07
Member #1
|
The standard measurements for stock that I use are as follows:
Measure Definition Calculation
Accuracy Actual quantity versus system quantity Actual quantity by SKU/Reported quantity by SKU
Damages Damages as a % of cost value Total damage/total inventory value
Days on hand Average sales days of inventory based on history Average inventory value/average daily sales (past month)
Storage utilisation Occupied space in warehouse as % of available storage capacity Space used/space available
Forecast accuracy Actual demand versus demand forecasted in a period % forecast accurate: Actual/Forecast sales per SKU.
Availability.
Quantity required versus quantity issued/delivered % available:
Ordered / Delivered
Per SKU.
As to specific measures pre automation..well I have never seen such guidelines..but I would expect that the sellers of such kit will have a view...this may of course require a "judgement" ...( sellers have a vested interests etc)!
Additionally would always counsel on any resultant standard figures on the above measurements.... as so very often "apples with apples" comparisons are not being made.
Also on any system...automated or otherwise......the performance is dependant on system design, package size, product mix, products availability,etc etc...
Best Wishes
Stuart
|
|
iomnet
|
|
|
System Administrator
Total Posts: 87
Joined: 12 Feb 07
Member #1
|
MRP folklore used to state that you needed an IRA of above 98% to run effectively with below around 95% being deemed as akin to chaos.
There are two main camps in terms of how one defines accuracy, there is a simplistic view that it is the number of counts that is measured so if I count 95 of a part when there is supposed to be 100 then that 1 count is scored as a miss or zero. Assuming that we had a count of 10 parts and 4 of them were as described above then we would state an IRA of 60%.
Others prefer to use a volumetric approach, especially when dealing with large quantities of cheaper parts. Assuming we have the 10 parts counted as above and each is supposed to have 100 on hand we have an overall volume of 1,000. now if we found just 980 parts when we carried out the physical we would assume that the overall IRA is 98%?
Possibly, but to be suitably harsh in how you measure it is far better to multiply the results. Lets say that of our four parts that are not equal to the supposed bin quantities the number counted is 97, 95, 90 & 98. If we multiply all our counts by percentage in a standard x times y times z string we get an overall IRA of 81% for that count.
Of all methods I recommend either the first hit/miss which is a good way to prove a point or the last method which is a more finely tuned version.
David M
|
|
iomnet
|
|
|
System Administrator
Total Posts: 87
Joined: 12 Feb 07
Member #1
|
The traditional view is of course that records should be 98% accurate for a system to be credible. The measurement of "98%" depends on the environment. An approach to this measurement is discussed in my book "Improving Inventory Accuracy" (Unashamed plug - but obtainable from the IOM Bookshop!!). In general the measure of accuracy does not need to be numerically exact, but sufficiently accurate not to impact customer service. In practice I favour an ABC based accuracy measurement.
Tony Wild
|
How accurate does stock need to be?
29 Jun 2007, 19:55