Supply Chain News: Inventory Metrics, Part 1

This column begins a 4-part series on inventory metrics.  Any organization involved with storage and handling of materials recognizes the importance of inventory accuracy.  It is one of the most basic of warehouse performance measures and can be applied to inventory in all its forms.   

In its classic form, inventory accuracy is a comparison of the physical count to what the system says is on hand either in total or by location.  For example if the warehouse management system indicates that 10 units of part number XYZ are in slot B0029, the inventory count accuracy indicates, depending on your choice of measurement criteria, either the percent of correctness or how frequently one can go to that location and find that the physical count matches the system’s on hand balance.  The formulas used by most companies are:

• Inventory count accuracy by dollars/units:   [1 – (the sum of the absolute variance in units or dollars/the sum of the total inventory in units or dollars)] * 100%.  This is a percent of correctness formula where the system inventory count or value is determined to be XX% correct.  For example, if the physical count was 167 and the system count was 175 the accuracy would be 1-(7/175) = 96%
• Inventory count accuracy by location:   [1 – (the sum of the number of locations containing an error/The total number of locations counted)] * 100%.  This is a frequency of correctness formula where the system likelihood that a system bin quantity is valid is XX%.  For example, if you count and validate the physical count of 250 locations and 247 of the locations were correct, the accuracy would be 1-(3/250) = 98.8%.

With the improvements in system tools and the increased use of RF based transactions companies have come to expect high levels of inventory accuracy.  While an inventory accuracy of 100% may seem unrealistic to some, the 2008 metrics study we just completed for the Warehouse Education Research Council reported benchmarks showing that median inventory accuracy levels by count or value are consistently above 99.6% with best in class companies reporting accuracy levels of 99.98%.  This holds true across DC’s in many different industries from manufacturing to retail.  The message here is that most companies have processes that are effective in controlling inventory. 

But the traditional calculations look at inventory as a snapshot in time, with no regard to the number of turns a part may have.  Basically - if you have a high level of inventory turns – then in theory - when you take a snapshot during a physical or cycle count at a specific point in time you could have actually managed MUCH more inventory than the on hand balance suggests. So we are seeing a trend by companies, especially ones with high inventory turns, to use a revised calculation for inventory accuracy that takes into consideration the total movement of material. The most common formula we are seeing is:

[1 – (# of net unit discrepancies / (on hand quantity+total# of parts shippedin the time period))]* 100

This formula is different from the traditional formula, in that it makes an attempt to look at the total unit throughput (based on units shipped) in the period (the time between counts) and it would allow for a higher number of discrepancies when you have a higher unit volume. 

For example:  On hand = 5000, Total = Total discrepancies = 10, Total shipped in period = 10000. The traditional calculation would result in an accuracy level of 99.8%, using the unit volume based approach the accuracy level would be 99.93%.  In other words the calculation is giving credit based on the increased volume. 

I see this as a good first step, to understanding the impact velocity has on accuracy, but it still ignores the actual impact a transaction has on inventory accuracy.

As an illustration, if on one extreme you have a totally stable inventory, there is stock in the warehouse, but nothing ever gets added to, or taken from it, you should be able to perform physical/cycle counts and have zero variance forever. On the other extreme you have a stock where you receive multiple replenishments during the day and pick parts by unit, every time you perform a transaction or visit the stocking location you risk an error.  A calculation that takes “touches” or opportunities for error into account may be a better way to measure inventory accuracy.  The accuracy calculation might be modified to read:

[1 – (# of net unit discrepancies / (on hand quantity+total# transactions)]* 100%

The number of transactions would include both receipts and withdrawals. 

Yet another possibility to consider is the nature of stocking units.  A distribution warehouse which receives and ships only full pallets may have a very high unit movement but again, very small transaction count.  And due in part to the difficulty of losing or misplacing a full pallet, accuracy tends to be somewhat higher.

Is considering volume in the accuracy calculation a positive thing?  Will it make the metric more meaningful?  What are your thoughts on this developing trend?   Do you think the use of different formulas could make comparison of performance across facilities or companies more challenging?

In the next three installments of the series on Inventory Metrics we will discuss practices employed to conduct physical inventory counts, review a couple of metrics that are used to help control inventory accuracy but are sometimes poorly tracked – damages and shrinkage, and finally focus on some metrics designed to help determine whether the inventory on hand is properly aligned with demand.

Agree or disgree with our expert's perspective? What would you add? Let us know your thoughts for publication in the SCDigest newsletter Feedback section, and on the web site. Upon request, comments will be posted with the respondents name or company withheld.