On Jan. 9 of this year, I posted a blog on “Best Practices for Online Surveys.” As I explained in that post, surveys can provide significant fodder for PR if handled correctly. I offered five helpful tips for anyone wanting to develop a survey.
This week I want to offer some additional wisdom from U.K.-based Stewart Rogers, the director of marketing technology at VB Insight, a leading and highly regarded source for news, analysis and reports on technology innovation. I queried him about his perspective on online surveys, and here’s what he had to say:
Is there an ideal survey length from your perspective?
Obviously, nobody wants to fill in a survey that has 50 questions and lasts 30 minutes, but there really isn’t an ideal survey length. As with most content, the key is to ensure the survey is interesting. The other aspect to keep in mind is variety; mixing up how you ask people questions and get their responses reduces fatigue. For example, I’ll use a dropdown menu for one question, a radio button for the next, and a slider for another.
How do you incent people to respond to your surveys?
At VB Insight we offer respondents the chance to see the data after we’ve finished the survey. Obviously, there is a big difference between raw data and the insights that we glean from that information, plus the additional data we use when creating studies and reports, but our respondents find the results of their survey interesting and useful. There are many ways to garner a response though—it really comes down to ensuring your target respondents are interested in the initial topic.
How do you go about pinpointing your demographic to participate in a survey?
It depends on the survey. Many of our studies are around marketing technology, so readers of VentureBeat are natural participants, but we also work with a range of data partners and surveying tools when needed. For example, in my Digital Pitchforks report—all about how people complain to brands on social media—we worked with Survata to poll over 11,000 U.S. consumers.
How do you figure out how many respondents you need to get a statistically valid sample?
You’d be surprised at the numbers required to generate a statistically valid sample. For a population of about 1 million people, you only need to survey 384 people to get a reliable answer. However, that’s for a single question. When you start asking for compound opinions, the requirements multiply accordingly. Of course, good studies don’t rely on a single survey to help generate useful insights, so we combine these results with other data to ensure we’re telling the whole story.
How do you determine when to distribute your sample?
We have a well-thought-out calendar of reports from VB Insight staff that will take us through to 2016, so we run our surveys based on a schedule that leads us to the publication date. This ensures we always have the freshest data to work from, which keeps our studies interesting and informative.
Hopefully Stewart’s insight, coupled with my original post, will provide a strong basis for planning your next survey. Keep us posted on your results!