# Company Survey: how to weight multiple units?

#### Darrin

##### New member
Hi,

It has been 30 years since I've taken a stats class and appreciate your input as I'm not quite sure how to approach this problem. For work, I'm about to do a customer survey to find out more info about how our devices are being used. I have about 700,000 devices out there. The customer base is quite varied with regards to the ownership of units. So hundreds of thousands of customers who have just one unit. We have 100 or so customers that 2,000+ units, and everything in between.

On my survey I want ask the how many devices each customer has but I'm little confused on how to weight this. One to one is straight forward if every customer has just one device. But lets say I get 20 customers with 100 units and 80 customers with one unit. How would you treat this? Would you consider this as (20 x 100 + 80 =) 2080 as the number of responses?

It just seems that either with or without weight the numbers are skewed. The goal is to have a 99% confidence. I guess if I survey 700 or so customers regardless of numbers it will be fairly accurate +/- 5 percent or so. Any thoughts, suggestions, formulas would be helpful? Again I appreciate the help.

Darrin

Last edited:

#### Subhotosh Khan

##### Super Moderator
Staff member
Hi,

It has been 30 years since I've taken a stats class and appreciate your input as I'm not quite sure how to approach this problem. For work, I'm about to do a customer survey to find out more info about how our devices are being used. I have about 700,000 devices out there. The customer base is quite varied with regards to the ownership of units. So hundreds of thousands of customers who have just one unit. We have 100 or so customers that 2,000+ units, and everything in between.

On my survey I want ask the how many devices each customer has but I'm little confused on how to weight this. One to one is straight forward if every customer has just one device. But lets say I get 20 customers with 100 units and 80 customers with one unit. How would you treat this? Would you consider this as (20 x 100 + 80 =) 280 as the number of responses?

It just seems that either with or without weight the numbers are skewed. The goal is to have a 99% confidence. I guess if I survey 700 or so customers regardless of numbers it will be fairly accurate +/- 5 percent or so. Any thoughts, suggestions, formulas would be helpful? Again I appreciate the help.

Darrin
That is not correct.20 x 100 + 80 = 2080

#### tkhunny

##### Moderator
Staff member
Why do you care how many devices? Why not just count customers?