Hello - a client of mine is launching a new internet data storage service. We are charged with validating the existing business model and doing some capacity planning (e.g. how many gigabytes of storage do they need to build into their network over time to meet customer demands)
We do not have much data available. However, from working with some similar product launches in the past, we are expecting that a histogram with storage units plotted on the X axis and number of of customers on the Y access would show the majority of customers with very low storage usage. In fact, because the company will offer a free trial, we expect the highest percentage of customers to have 0 data stored. Only a very small fraction will be towards the high end.
We do know the minimum and maximum expected storage. It ranges from 0 to 150 (units are not important)
I am looking to apply a statistical model to this problem and need some input. Given a storage level, I would like to be able to calculate the cumulative probability density function for that level of usage. In other words if someone says "how many customers are using <25storage units" I can enter 25 into an equation and it will tell me what % fit this profile.
We have very limited data available. all we know is the total number of customers expected and the min & max storage plus the general assumption that the probability function will be skewed towards 0.
This seems arbitrary and I would love to have time to sample the customer base or gather more information. But we are time constrained. Are there any thoughts on what type of distribution we could use? I would assume people have had a similar problem and that perhaps there is a standard distribution for modeling customer purchase behavior or usage. Our plan is to pick a model, document our assumptions, and validate it as more data becomes available - then revise the model.
PS I am a six sigma black belt but about 5 years out of practice. I apologize if this is a very simple question and not worthy of this forum! I need to brush up!
Advertise on Talk Stats