Subscribe Bookmark
louv

Staff

Joined:

Jun 23, 2011

JMP for Data-Driven Decisions at Home: Cell Phone Use

A while back, I was surprised to receive a huge cell phone bill. My family had greatly exceeded our monthly allotment of minutes. In response to that episode I became extremely conservative with my cell-phone minutes allotment that I purchased. Then as time passed I decided it was time to analyze my family’s cell phone usage data in JMP to more correctly assess the situation.


Initially, I visualized the distribution of 24 months of cell phone minutes usage for a family-of-four plan. I could see that the data was normally distributed and that 1,400 minutes looked like a reasonable amount of minutes for us to purchase based upon the available plans.




Digging deeper into the data, I thought it would be prudent to add a time element to this data and take advantage of the Control Chart platform in JMP. I wanted to see whether any trends were developing before making a change in my contract.


This analysis showed that the process appears to be in control, residing between the lower control limit (LCL in the Control Chart below) and upper control limit (UCL). In addition, it appears that 1,400 minutes would be a good choice for our minutes per month because it is slightly outside of Zone B in the Control Chart. This means that roughly 95% of all of my data in the future, if my process remains in control, should reside within Zone B. I concluded that this is the level of risk that I can tolerate for a $20 savings each month by cutting back our usage to 1,400 minutes per month.




But wait! There were still more discoveries to make! I peeled the onion a little further on my process. By sorting my minutes usage by phone number and stacking the data, I used the Phase feature in the Control chart platform. This allows me to visually see the usage by user.




Clearly, although the entire family is a process that is in control, as shown previously, the components of that process show interesting usage patterns. In the name of family harmony, I choose not to elaborate on those conclusions any further. Suffice it to say that I do not have many friends. However, I am happy to see that I have the lowest variation, and the process with respect to my phone usage is Six Sigma.

5 Comments
Community Member

Kun wrote:

Very interesting example. Can you send me your JMP spreadsheet in this study, so that i can try by myself.

Kun, from China

Community Member

Charles wrote:

Will you please send me some examples? My e-mail is ckeating01@yahoo.com.

Thank you!

Community Member

Lou Valente wrote:

You ask a good question. Since I have 28 years of performing Design of Experiments for my previous employer, I can tell you that without exception the DOE approach is the most efficient and effective way to experiment. As far as your question around running a DOE to optimize a Web site, for example, I can send you some examples of how DOE is used exactly for that.

If you own a copy of JMP, you will find a very helpful tutorial located at the Help pull-down menu that takes you through the DOE process step by step. If you do not own a copy of JMP, please download the 30-day trial and check it out. JMP and DOE certainly made a significant difference for me with respect to the efficiency and effectiveness of my experimentation.

Community Member

Charles wrote:

Thanks for the post. By the way, I read an article on experimental design that is quite unlike the "conjoint like" analysis that was provided several weeks ago. Instead, it was a factorial design that was used to determine the most criterial or levels of various web banner advertisements. For example, a company might have a combination of sizes, messages, offers and prices. How would this be done using the experimental design platform in JMP?

I hope you can understand my question.

Thank you

Community Member

Daniel wrote:

Excellent! I've been looking for an easy-to-understand example using JMP's control chart platform before diving in with some of my own data. This was tremendously helpful. I may want to do the cell phone analysis too!

Article Tags