Technically, you should base the size of the sample on the need for

credibility and the risk level involved.

There are websites you can use to determine error margin calculations.

But first, you've gotta look at that purpose.

You may have heard before that you need at least 400 people for an internet survey.

How do people come up with the number that is tossed around quite a bit.

Here's how.

Market research commonly aims to get an error margin of plus or

minus 5% at 95% confidence.

That is generally the industry standard across the board.

In order for you to do that, you need to have something like 387 but

people always say 400.

That's why you often hear 400 for a sample size.

That's a general sample size that you need to have in order to be scientifically

credible among the profession.

However, people often sample fewer, say 50 or

100 and they accept higher levels of error.

But the risk may not be as bad.

There are other cases where people will survey 1,000 or 5,000.

That may be to reduce the error margin, but

it's also to increase public perception.

Every news outlet that does a pole never reports 400 because 400 is not

a cool number.

1,000 is what they report, they did 1,000 surveys of registered voters or

a sample size of 5,000 registered voters.

Nobody announces 400 on national television.

When you survey 1,000, you have an error margin of plus or minus 3%.

It's very unlikely that you're going to be wrong.

There's still that 3%, but your chances of being wrong are less.

It's also considered newsworthy to have 1,000 or 5,000 for the media.

So aim for a 5% error margin or less to be considered credible.

There are many margin of error calculators or sample size calculators.

The one I use is from the American Research group.

But there are many others including ones from SurveyMonkey, Qualtrics, or Raosoft.

Using such calculator, you can insert a sample size of say 400 and

this assumes a very large population and it'll produce a error margin of plus or

minus 3% and 95% confidence.

And this assumes a population of the U.S.

or large populations in California or Pennsylvania.

However, when you have a customer base of, say, 643,

you're sample size doesn't have to be 400.

It's going to be something different.

If I were to use a sample size calculator and plugged in that I had a population

of 643 customers and I want a plus or minus 5% I would run the calculation,

and a calculator would tell me to do 240 surveys of my 643 to get plus or minus 5%.

The American research group that provided the calculator I use is a company that

does surveys and survey research.

This is just one of their tools to get you onto their site.

You can also do the same thing with SurveyMonkey or

contracts who also have sample such calculators.

Now let's consider sampling,what kind of sampling do

you use from the many available.

A common tool for most market researchers is the random sample.

This is dine in a manner to exclude any bias or

systematic errors in selecting a sample.

It's typically the best way to sample, but also more rigid.

This approach has a higher likelihood to get market research results

that represent the population.

However, there are other sampling methods as well, such as a convenience sample,

where the researcher gets the first x number of people they need to

address a situation.

There are clear disadvantages to this, but

the major advantage of this approach is time.

Most people who employ a convenient sample do so in a low risk situation.

There are other sampling methods like a quota sample

where you're done when you reach a certain numbers of quota.

There's also a snowball sample where you build your sample based on your existing

sample or participants.

The ladder is often acceptable approach when you're having difficulty finding

qualified research participants.

Some people use this approach for in-depth or

executive research sometimes called key informant or an opinion leader research.

There are many other ways of sample and entire courses are books

dedicated the sampling, explore the subject in much greater detail.

A random sample is the preferred method of sampling, however,

there are many corporations, companies,

organizations in market research firms that use other sampling methods.

It really comes down to what their constraints are and what their goals are.

If their goal is to get quick information,

such as talking pandas versus talking rabbits,

they may not need a random sample for that, they may just need a tie breaker.

In this case they might just want to do a quota sample.

Let me explain quota sampling.

That was the example I gave earlier on the talking pandas and

talking rabbits situation.

In that the researcher needed to get the first 50 people they can get from

a target age audience of those under 25.

A random sample, you actually have to say, okay,

there's ways to create randomness in terms of how to find these first 50 people.

You take a random number and you take that 17th person and

then you do another random number and you take the 212th person.

You use a random number generator, which uses some kind of algorithm

that gives you random numbers to choose emails or phone numbers.

For the quota sample, there's no randomness about it.

The quota sample and a convenience sample are actually a lot alike.

It's all about ways to get to that number you need for

your sample size very quickly.

A snowball sample, on the other hand, is used when you have very few participants

and that are hard to reach participants.