About a year ago at work, I was brought into a customer service project that involved reviewing phone discussions with customers, and rating the service level based on defined criteria. My contribution to the project was to be towards the end of the project, reporting on the results and combining the survey data outside of our HR system with data from or HR system to tie the results to different departments and employees. I was brought into the project a little earlier than expected because they were trying to decide the best way to get a start on the project and get a handle on the data.
Once the phone records were examined, they found thousands and thousands of conversations, with no real rhyme of reason to them. They were all saved digitally, but without any particular clue as to the context of the conversation. Also, there was no way to tell from the file names which department generated the call.
After some initial discussions and ideas, I proposed that a random sampling of them would likely be the best way to statistically review the information. I did some initial research (unfortunately, I had lost the memories of how to do it from my previous stat classes) and found what number we would have to sample in order to get a statistically sound representation of all calls. I used Slovin’s formula discussed here: https://www.statisticshowto.com/how-to-use-slovins-formula/. I was able to determine we need around 360 calls for our sample.
We then decided to assign a number to each call, then use a random number generator to pick the 325 sample calls. Once that was picked, the committee that would review the calls would then be assigned a fair number of them, based on the length of each call. Some were as short as a minute, while others were 30+ minutes. We would assign the calls so that all would have the same amount of calls in minutes.
Unfortunately, that is as far as this project has come. Covid forced us to work from home, and this project has been delayed. However, it did spur my interest in statistics again, and is a part of the reason I’m taking classes.
Stephanie Glen. “Sample Size in Statistics (How to Find it): Excel, Cochran’s Formula, General Tips” From StatisticsHowTo.com: Elementary Statistics for the rest of us! https://www.statisticshowto.com/probability-and-statistics/find-sample-size/
Chapter 4 material was very interesting for me, as I was able to start understanding most of the concept of statistics that I use to struggle with back in high school. Some of the concepts and tools taught this week are population and samples, normal distribution, probabilities, measures of centralness, central limit theorem (CLT), hypothesis testing, etc.
And as a visual learner, when dealing with a very large dataset, the tools that I like and have used in the past are tables and graphs found in distribution. Often, tables and graphs help us make a better sense of the data given. In one of the previous jobs I had, every week I would monitor and collect repair cases (incoming, pending UM decision, resolved, not resolved etc.) for our warehouse using a built-in template in excel to track repairs and include any comments that will help upper management (UM) make critical decisions based on available data from the warehouse. The report would include a table showing the number of devices already in the warehouse as well as their status, those coming in for the first, second and third time and a visual histogram would include the distribution of those devices based on their repair status.
UM would only focus on cases with the status ‘pending UM decision’ as those were the ones coming back to the warehouse for the third time and based on how frequently the same issue was occurring and how long on average the issue was happening after the second repair, they would offer a free replacement to the customer to avoid bad business and do further testing on the defective devices, then collect data that will help them improve the performance of their potential future devices.
In the end, when used the right way, these concepts and tools not only can help us answer questions that arise in real situation, but they also lead to powerful results if we rely on them.
- Bell, P., & Zaric, G. (2013). Analytics for managers: With Excel. New York, NY: Routledge.
- Kernan, D. (2007). Natural Resources Biometrics. Retrieved from https://courses.lumenlearning.com/suny-natural-resources-biometrics/chapter/chapter-1-descriptive-statistics-and-the-normal-distribution/