How to determine survey sample size

"This may be slightly contentious, but I'd like to think that this is a practical way to determine what sample size you need for your next market research project."

We'll all have come across calculators for determining sample size, based on confidence levels and confidence intervals. They are readily available in text books and even more accessible now online. We continue to use them, and they are indeed valuable tools. This however, is a slightly different approach, and perhaps not for purists, but it maybe a more practical way forward.

The approach I am suggesting is based predominantly on what your business can afford.

Market research, like any other input to the marketing mix, is a business-critical activity, but it must be affordable. There is no point doing something that is beyond your budget, so we have to be sensible and practical, when deciding what to do, and also what not to do.

Now I'm not suggesting for one minute that your sample size should be one person. Often I've heard people say they have done their "own research", by asking their own mum their opinion, or they've chatted to an old friend over a coffee. Clearly, there is nothing wrong with asking the opinions of others, particularly people whose opinions you respect. What we are talking about here is primary market research, which is a little bit different. So, putting to one side for a moment the idea of asking just one or two people, what I am talking about here is larger-scale research, but still at an affordable level.

For any research I would argue that even relatively small numbers of people still have their role. Here I'm not just talking about just one or two respondents. No, but I firmly believe that if you ask even a small number of the right sort of people from your target market, or from your own customers, there is still great value to be had in conducting that sort of market research exercise.

But what is the right number? Again, I'd argue it's what you can afford. So, conducting a survey amongst 25 people if you are operating a small business will certainly have some benefit. I don't think you necessarily want to be making particularly life-changing business decisions on the views of two dozen people, but even so, this will give you some indications, perhaps backing up a hunch you have. And looking at this another way, plenty of businesses have used 3 or 4 focus groups (which is a similar number of people) for some pretty big decisions. True.

I would argue that if you can run a survey perhaps amongst just 100 or so people, provided you think carefully about who you speak to and how you speak to them and about what; and crucially, if you can ensure they are representative of your market, then there's incredible value to be gained from even a relatively small-scale survey.

Going up in scale slightly, I would tend to use a rule of thumb based on the number of different segments or cells. This approach gets you to think about key segments or groups that you would like to compare and contrast the results or opinions for. So, one example might be 'frequent users' vs. 'infrequent users'. Another could be males vs. females. Taking this approach I would use as my ideal number 100 per cell or segment. So, in the latter case if we wanted to compare the responses of males and females in our survey, I'd recommend at least a 200 sample survey, with 100 in each cell (i.e. 100 males plus 100 females). The same works for other cells or segments.

Following this logic, if you wanted to examine behavioural differences between age cohorts, and assuming six cohorts (e.g. those aged 18-24 years, 25-34, 35-44, 45-54, 55-64 and 65 years+) this therefore gives us a total sample size of 6 x 100 = 600. If my budget was really tight, but I still wanted to compare these six age segments, at a real stretch, I'd be prepared to drop the number per cell down to a bare minimum of 50, so 6 x 50 gives us a sample here of 300. This won't be ideal, and some statisticians would probably disagree with me, but this approach is all about being practical.

At the top end of the scale, a lot of national brands would be more than happy with a 1,000 sample survey, provided the respondents were representative of their population or target market. Clearly, there is no upper limit on your sample size and, if you have the money, who's to stop you running 2,000 sample surveys? However, for smaller businesses, or regional brands, the more sensible approach maybe the pragmatic approach.

In summary, I'd say spend what you can afford, even if that seems to be less than wanted, there is nothing wrong with smaller sample surveys.

There is still value and insight to be gained, but please still be very careful of the infamous "sample of one".

About the Author

Paul Latimer is an experienced market researcher, having held a number of marketing roles in a wide variety of industries he moved over to focus on market research from the agency side in 2004, before setting up his own market research consultancy, Latimer Appleby, in 2012. He is a graduate of the University of Birmingham, and holds several diplomas including those from the Chartered Institute of Marketing (CIM), the Market Research Society (MRS) and the Institute of Data & Marketing. He is a Certified Member of the MRS, and a Fellow of the CIM.