Please do not tell RG the names of the companies or what they do.

Is your population 300? So will you take your results and try to apply them to the missing 20? This is a bad idea.

Is your population 280? Will your results be used to describe the current state of these companies, or will your results try to inform about the future state of these or other companies that might enter this market or similar companies in other locations? If the former, then the population is 280. If the latter, then you have to do a bit more assuming/hypothesizing/guessing.

280 is a small population. I would try for a census and sample all 280. You will probably not get all 280 surveys back even if each survey has only one question. Keep track of those that return the survey quickly, versus those that you have to call 3 times to get the survey back. This might help you figure out something about the non-responses.

The fewer responses you get back, the more uncertain you will be in how representative your responses are of the entire population. Your sample size relative to the number of questions may limit your options for analyzing the data.

Thank you! The population is 300 and I thought about making the 280 the sample size - so sending the survey to the 280.

I wasn't sure if the 280 should be the population and using a selection of those as my sample? I am afraid however of too few responses so your comment of surveying all 280 is very helpful.

If you select the 280 based on software version, then the remaining 20 are not part of your population. To take 280 surveys in a population of 300 you would need to randomly select the 20 companies that do not get the survey out of all 300. As-is, the 20 are an unsampled subpopulation and your survey results would not apply to them. Another option is to try for all 300 companies, and realize that the 20 that have the other software might have to be treated differently in the data analysis. However, you won't know that until you have the data and try to analyze it.

Dear 280 will be your population under study 20 will be included because they do not use the software you are working for. Randomly select 70 companies that will be your sample size sufficient enough to draw a conclusion. Or other option is to study the whole population because it is not too large that you cannot cover and many of them will also be dropped out as non-respondent . Another aspect is financial one that also is clear in your study because it is a sort of profarma filling that will not cost any thing.

My personal advise is to cover all 280 and analyse the responders out of them that will be much stronger study. Actually sample size is selected where population is very large that one that cannot be covered or it will take too much time, human and financial resources.

If any other query just ask we will help you at maximum ?

While 70 is 25% of 280, that is not a good justification for suggesting 70 surveys given the information provided. We don't know the size of the survey (how many questions) and we don't know anything about the survey. The company will not fill out the survey. It will be a person in the company. Depending on the questions, there may be differences between males and females answering the questions. 70 surveys has now become 35 replicates. There might be two size brackets (companies with 5 or fewer employees, companies with more). You now have 17 to 18 replicates. This assumes an even distribution in categories, and that is unlikely. This assumes that everyone returns a survey. Again unlikely.

We don't know the type of analysis. Will this be PCA, regression, or univariate models using a multiple compariosn test?

The only rational plan here is to try for all 280. Reviewers may accept 70 returned surveys out of 280 tries, though you might get criticism from the low response rate and questions about how relevant the results are to the total population. However, this will be much better than trying for 70 and getting 18 surveys from a 25% response rate. Given the small population size, the closer you are to a census (280 returned surveys) the easier it will be to get this published in a peer reviewed journal.

There are 280 companies that provide a specific version of the service that you are researching. Is the service provided by these companies identical? Or do they vary and if so how would you measure or rank this variation? In other words what are you measuring that you would conduct statistics upon? As an example, if you were sending a rock sample to various labs for age determination and you wanted to get a range of possible ages, and you wanted to minimize the costs of these tests, a sample size of 30 would be good. But if there were several different methods for age determinations, then increased sample size would be necessary if you were to compare different techniques as well as differences between labs. Without knowing your sampling frame and the distribution of your measurements, adequate sample size can not be estimated.

Hi, the service is identical and I want to see if they hold the same or different opinions of a customer group

Kirsten, good. Then let's look at the possible variance around opinions. How many opinions can there be and again how would you measure or rank them? If it is a simple 1 or 2, then sample size can be small. As the number of opinions increases, then sample size should also be increased. For a small number of opinions, you could start with a small sample size, summarize the results and do your statistical testing. For instance if you got 10 similar opinions out of 10, that might satisfy your analysis. You have no variance with a sample size of 10. That would be like getting 10 heads in a row from a coin toss. If the first toss is not specified, then this is equal to 2 to the 9th power, or 1/512. As the results become closer to 50/50 your variance increases so sample size must increase to get meaningful confidence bounds. To maximize your analysis and minimize costs, you might start with an open ended sampling plan and end the sampling when your results are adequate for your study. For cost estimates prior to the start of the study, you could estimate sample sizes based on maximum variance, but have the option to stop the study should you encounter lower variance and and are obtaining adequate results. On the other side, if opinions can be quite variable or even open ended and costs are of no consequence, then it might be necessary to conduct a complete census and show the complete results.