Tips: getting a cost-effective sample size

What is the best trade-off between statistical accuracy and cost?  Do I need to do 1,000 surveys for the results to be reliable?  Usually not.


You can reduce your margin of error with a higher sample but that there are diminishing returns to increasing the sample.  So, if you increased your sample from 100 to 200, you bring down the margin of error by 2.9% (good). But if you increase your sample from 500 to 1,000, you bring down the margin of error by only 1.3% (not so great!).


A higher sample may be needed if the population being measured is not homogenous on some factor or factors.  For example, political polls need high samples because there may be real differences in voting intentions on a range of factors (e.g. gender, age, region of country, whether living in urban or rural area, etc.)


Tip: Decide whether your market can be considered homogenous or whether attitudes and behaviour may differ on a factor.  For example, if your market differs based on gender, you could do 200 surveys of men and 200 of women.


Note: margins of error are for an answer of 50% at the 95% confidence level.