I am frequently asked by clients, “what sample size do we need”?  In a previous blog I redirected this question to focus on the importance of response rate.  But here I’m going to tackle the question of sample size.

Recently I’ve discovered a handy table (Thank you Survey Monkey) for calculating the sample size required at three different error levels for various population sizes.  Okay, let me explain that last statement.  In a lot of survey research we are using a sample to make projections to a very large population e.g. all Canadian adults.  In this situation we can readily calculate a margin of error for different sample sizes.  However, when we’re dealing with a finite population e.g. members of an association, we need to use a slightly different formula to calculate the margin of error.  In a nutshell, when my total population decreases, the sample size required for a specific margin of error also decreases.

The following link will take you to the aforementioned table.  http://svy.mk/onjQdE

As an example, let’s say you have an association with 1,000 members. You are contemplating introducing new benefits for your members but you want some feedback on their relative appeal.

You want to be very certain that you are making the right decision so you decide to set your margin of error at +/- 3%. This means that if 50% of your respondents say that they like benefit “x” then we can be reasonably sure (19 times out of 20) of being accurate within +/- 3%.  This means that a total census of members would reveal an answer of not less than 47% and not more than 53%

In order to achieve this level of accuracy you will need to interview over half of your membership (525 members to be exact). That’s a response rate of over 50%. Achieving that level of response can be tough. We’ll be talking about this in future blog postings.