I have been asked about statistical inventory sampling so many times that I felt that I should share my thoughts in an article (or many). In simple terms, statistical sampling is the selection of a subset of items to represent a population and it is a very useful tool for establishing inventory accuracy.
In order for the process to be considered representative, there are a few fundamental rules:
- Each item in the population must have an equal chance of being selected (i.e. no targetting or cherry-picking).
- The selection of items to sample must be random and free of bias.
The following are some common terms you would encounter: confidence level, sample size, population size, margin of error. The vast majority of the time, the confidence level and margin of error will be dictated to us, as this determines the precision and robustness of our results. A very common set of parameters typical in DoD supply chain policy is 95% confidence level with a +/- 2.5% margin of error.
If talking about confidence levels and margins of error is still fuzzy, fear not, there are tools that make the whole thing practically plug-and-play. I even developed my own spreadsheet calculator in one of my articles on my website (link here).
Note: if you, like me, are “of a certain age” in Navy Logistics terms, you might remember a program called STATMAN that for many years was the Navy’s standard tool for these calculations when dealing with inventory accuracy.
In order to conduct what is called “Simple Random Sampling” there are only a few computations that we need to make: 1) the sample size, and 2) the accuracy results (usually a percentage or ratio). We were going to compute the accuracy anyway regardless of method, so that leaves the sample size as the only new thing we need to figure out. To get the sample size, simply plug the following parameters into the spreadsheet calculator linked in the previous paragraph (or an online tool like this one): 1) population size, 2) confidence level, and 3) margin of error. This will give you the minimum sample size that needs to be collected in order to satisfy the parameters. Select that number of items at random from the population – you can do this manually or using most spreadsheet programs. Then conduct the inventory and compute the accuracy percentage. Whatever percentage you get (e.g. 93%) would be expressed in terms of a “confidence interval”, which is nothing more than stating the margin of error with our results (e.g. 93% +/- 2.5%). That’s it!
Now, like anything, it can get much more complicated. There are sampling strategies that involve stratification, different allocation methods, etc. that are well beyond the scope of this article. Therefore, we will leave things here for now and I hope that you found this article useful.
If you would like to dive a little deeper into the statistics involved in this type of sampling, there is an excellent video in the zedstatistics channel in YouTube titled “What are confidence intervals? Actually” that I highly recommend.
References and further reading:
- GAO’s Using Statistical Sampling, publication PEMD-10.1.6
- Guy, D. M., Carmichael, D. R., & Whittington, O. R. (1994). Audit sampling: an introduction to statistical sampling in auditing. Wiley.
- Cochran, W. G. (1977). Sampling techniques. John Wiley & Sons.
- COAN904, Probability Sampling Course, GAO. (Note: sadly, the course no longer appears to be offered externally, but the manual can still be found here)