Running a survey? Tips to get started.

Market research often involves using survey tools like SurveyMonkey because it's cheaper than focus groups, doesn't involve travel, and uses what you may already have – email addresses for your target group. But just because it's easy doesn't mean you can't screw it up – equally easily.

1 Don't ask too many questions. Identify the core of what you need to know.

Surveys that are too long are unlikely to be finished by enough people to make the results meaningful.

2 Decide on the criteria for acceptable participants and screen out those who don't meet the criteria early.

If you only want the opinion of those currently performing a particular role or using a particular service, make sure you check that early on – in a question as well as in your distribution messaging. This way people don't begin the survey and then realise they're ineligible.

3 Allow an option for a nil answer, e.g. I don't buy textbooks / dresses / meat.

This ensures that your results won't be contaminated with false positives. If your lowest option is "1 to 2 times" or "every month" (instead of "0 times" or "I don't buy this at all"), respondents who don't buy / use any at all will probably choose the lowest option, despite it being inaccurate.

4 Ensure your answer categories are contiguous, ie. there are no gaps between them, and no overlaps.

So for the question "How many times a week do you eat eggs?" don't do this: 

    A. One to two times [What happened to "0 times" for those who don't eat eggs?]

    B. Four to five times [What happened to "three times"?]

    C. Seven times [What happened to "six times"?]

    D. Seven times or more [Overlaps with option C]

Rather do this: How many times a week do you eat eggs?

    A. I don't eat eggs at all [Allows a nil answer]

    B. One to two times [No gaps]

    C. Three to five times [No gaps]

    D. Six or more times [No gaps]

5. Do a trial run of your survey with a diverse subset of your target group.

This should highlight issues with the survey setup, which you can then update and push to the actual target group. But remember: if you update the survey, you won't be able to include data from the trial group in the final data because the survey will be slightly different. But it's better to have good data.

If all this seems a lot of effort and a bit fiddly, ask an editor familiar with survey structure to help you with the setup. They could even help analyse the data, compile charts, and write the report.

I used SurveyMonkey repeatedly for market research while I was at Oxford University Press SA, mainly quizzing teachers about what they wanted to see in new dictionary and literature products. It was a useful research component especially when paired with other methods that allow a wider reach (participants who may not have phone data to use for this purpose) and a deeper reach (in-person or phone interviews of some length). But survey data is only as good as the survey setup.