Sample size: How many people should take the survey?

in Survey tips
on Wednesday January 29, 2014 @ 3:42

The sample size, or number of people you need to complete your survey, depends on two factors – how big is the population and how confident do you want to be in the results.

Morning rush hour at Baker Street underground station

Most people are surprised to find out that the number is smaller than they anticipated. I was working for a young company in Estonia and we had recruited a local market research agency to track our brand progress. Being ignorant about market research we were ever suspicious about the fact that they were drawing conclusions from a nationally representative sample size of 1000.

The sample size of “only” 1000 people simply didn’t feel big enough to make authoritative claims for the whole population of a whopping 1.4 million people.

The better news – it’s easy to figure out the sample size needed for your survey. All you need to know is the target population size – the number of people relevant to your study.

Couple of tongue-in-cheek examples:

Population Required sample size
528 - people who have been to space 223
10,490 - athletes at the London 2012 Olympic Games 371
110,000 - wine growers in France 383
5,300,000 - all Hebrew speakers 384
50,000,000 - everyone who’s bought Michael Jackson’s "Thriller" 384
1,344,130,000 - everyone in China 384

Visitors at the British Museum, London

Why go bigger than that?

Hang on! Why are some surveys then conducted with panels in the tens of thousands? Two main reasons:

  • Confidence - sometimes you want to be 99% sure, instead of the industry standard 95%. This will immediately increase the required sample size.
  • Getting valid data from questions that are applicable only to a sub-set of the respondents. This becomes important if you branch off respondents based on their answers.

An example of the latter: A study among iPhone users asks "Have you used in-app purchases?" and those who answer "Yes", are required to answer 5 more questions on the topic. If you started with 400 respondents then maybe half will answer those extra questions. Will 200 responses still give you valid data?

There are easy to use sample size calculators available on the internet but this table below gives a pretty good overview to kick things off.

Population size ±3% ±5% ±10%
 500  345  220  80
 1,000  525  285  90
 3,000  810  350  100
 5,000  910  370  100
 10,000  1,000  400  100
 100,000  1,100  400  100
 1,000,000  1,100  400  100
 10,000,000  1,100  400  100

It's always good to consult with an expert if in doubt. Our client services team is always happy to help out when it comes to mobile-specific research, sampling questions etc. Just get in touch »

Join our Newsletter