It is impossible to survey every single customer you have. So, how do you know the sample you chose will give you the correct result?

Thanks 

No, it is impossible, but you can get a sample out of them and ask them specific and clear questions In order to get correct results 

the good result will come when more than one man work

If time / cost is limited, do a sample and state that you have taken a sample reading. Take a few highest risk cases, some lowest risk and a few in the middle but don't do them all. 

1.You can use random sampling

2.you can use the frequency of their visit to the shops

3.You can use their taste of product or services

4.You can use the class of your consumers too.

by segmentation methodology. 

Thanks for invitation,

In a very prcise wording via "Segmentation Methodology".  

thanks for your invitation 

i go with experts answer

This depends on the number of customers and medium of the survey, an email or SMS survey may give you a massive capacity but it would be limited in scope, a live or IVR call on the other hand may have a much lower capacity, it could be either impossible or too expensive which may prove to be a challenge in many cases especially if you are in a B2C business model, the best way is to apply segmentation to your customer base, this may vary depending on the industry and objective of the survey, i.e: what you are trying to find out, once you have your customer segments you can follow a random sample as stated by Abdelrahman, but from each segment, if you have specific doubts about the results, you may repeat the survey using a different sample to confirm the results.

It is very much possible, depending on ones effort put in the survey and management of the periodic time. For accurate results, random sampling and surveying customers at an almost the same distace apart would also help

You need limitation on either criteria or random numerical value. if you choose 200 shoppers to take your survey. 

For instance, if you are looking for middle age and young shoppers, you can limit people who choose age range > 40 and < 18. 

You can also eliminate shoppers with a numerical value, if you give shoppers serial numbers 001 - 200 you can choose to eliminate shoppers with 5 in their singular serial number, (5, 15, 25, 35, ...etc) 

These limitations need to be added before the results, so you can get a better and more accurate results.