[sk8ramp]

In an effort to get relevant market research for a new business plan I put a questionnaire on my sites, all the sites we advertise on and some relevant forums. My question is how many people constitute a fair market analysis? As investors read the plan how many people would a survey have to question for them to put any weight in the results?

[jasaunders]

Depends on the type of survey, question, format, and tolerance for error. Some surveys may be acceptable with 10 responses, while some may need 1000 to be credible.

[sk8ramp]

OK thanks, I'll try to be more specific. As you can probably tell by my nym and avatar it is pertaining to skateboarders and their tendencies. The research has to do with how often the average skateboarder uses a specific service offered to them. It has been posed to avid skateboarders and asks very specific questions. Does that narrow it down and make the question easier to answer?

[James Strachan]

I would say that the more people you ask or survey the better and more accurate you result will be.
It all comes down to numbers people.

[ADVERTISE HERE!]

Really, you need at least 30 to have any kind of statistical significance. More than numbers of responses though, I would focus on writing the questionnaire that effectively elicits the most useful information possible - a mistake that the MAJORITY of questionnaire writers make.

[jayrobinson15]

You do need at least 30, but I would say to send out at least 100, because you will not get complete responses, or maybe any response at all, from a certain percentage of people. Plus if you do get them all back, 100 will give you a much more accurate picture of the results.

[ScottRobertson]

I suggest using AT LEAST 100 COMPLETE/ACCEPTABLE responses for any meaningful statistics.

[ADVERTISE HERE!]

Originally Posted by ScottRobertson

I suggest using AT LEAST 100 COMPLETE/ACCEPTABLE responses for any meaningful statistics.

Why? Where does the number 100 come from? Any statistically-significant reason? Or a shot in the dark?

[ScottRobertson]

The number 100 has nothing significant about it... 95+ if you would prefer. 100 is a good number for making it more simple for someone who clearly does not have much statistical knowledge.

But using anything less (such as your suggestion of 30) is not a good idea - it does not give enough opportunity to find true results - I believe that 30 responses would be too prone to error and I do not think that the OP has any knowledge of statistical devices such as a samples margin of error calculations/uses.

Why do you believe 30 is the magic number?

[ADVERTISE HERE!]

30 is the number necessary to reach a statistically significant conclusion at the 95% confidence level. Again, any statistical reason for 100? There's really very little advantage to 100 versus 30, so I'm curious as to why you would demand 100 rather than being ok with at least 30. Statistical confidence is the only thing that matters.

[sk8ramp]

Well i am up to 200 now so I guess I am in the clear, thanks for the input folks.

[ScottRobertson]

Originally Posted by ADVERTISE HERE!

30 is the number necessary to reach a statistically significant conclusion at the 95% confidence level. Again, any statistical reason for 100? There's really very little advantage to 100 versus 30, so I'm curious as to why you would demand 100 rather than being ok with at least 30. Statistical confidence is the only thing that matters.

Stat confidence is not the only thing that matters. As confidence increases, the margin of error also gets wider. It is obviously important to get a high level of confidence (90+ [95 generally being the standard]) Also, the number of responses does not determine what your statistical confidence will be .. you decide what you want your confidence level to be and get your sample size from there.

Heres my math:
Because we do not know what the OPs' questionnaire's proportion for success we have to be conservative and use .50 (this would give a sample size that is greater than needed.

With 30 respondents:

Std Error=.09
z*=1.96 for 95% confidence interval

Therefore margin of error = .18

With that you can say that 32% - 68% of his respondents will be a success.

With 100 respondents:

Std Error= .05
z*=1.96 for 95% CI

Margin of error = .10

~~

With your 200 respondents (assuming you mean 200 usable ones) your margin of error is .07 with a 95% confidence interval. I think that

Finally, for anyone that doesn't know what any of these terms mean, please feel free to PM me and ask or ask here. I was going to define them within my post but am too lazy, haha.