Portland Marketing Analytics (PortMA) | Portland, Maine

How Many Responses Do I Need?

by  

http://flic.kr/p/4StwRYI often get asked how many survey responses you need when surveying event patrons.  Yesterday I was speaking with one of our public sector Clients about this very thing.  The main questions were… do we collect more surveys at less events or less surveys at more events?  And how many surveys should we collect at any event?

Great questions.  When it came to how many events to target here is what I said:

It’s always better to collect less surveys at more events.  Any data collection effort should focus on taking the things that could bias your results (and there are a ton of them) and making them as random as possible.  I don’t want a Brand Ambassador who is on top of the game 99% of the time but happen to be having a bad day skew the results.  If there was a storm and they had to shut down the event set for a couple hours, I don’t the short-term line that forms as they set things back up to be overly reflective in the survey results.

Spread out your data collection.  Let these things get burred in the data so the true impact can come out.

What about the number of survey responses.  There is a clear guideline here as well:

When comparing things, you want each “thing” to have at least 30 responses (more is better).  A thing might be all retail events vs festivals.  Or it might be all Texas events versus So. California.  Regardless, if you have less than 30 responses, the algebra falls apart.  More is better but 30 is the minimum.  Therefore, I tell folks that if you got at least 30 per event than you can’t go wrong.  Do more if you can but hold that as a minimum.

If you want to get a clear understanding of the big picture… you want those string of 30+ per event to add up to 400 or more across multiple events.  Knowing the big picture is knowing that you surveyed enough people that if those you didn’t survey had a voice, it wouldn’t change your interpretation of the results.  This is where phrases like “generalizing to a population” or “statistical significance with 95% confidence, plus/ minus 5% margin of error” come in.  You don’t have to worry about that.  Just make sure you have 400 or more in your recap and you’ll be fine.  Doesn’t mean you can’t do a lot with less than that but 400 is a good target.

Those are our guidelines.  You take good reserach practices, mix them with budgets getting presure from every angle and you make the compromises you have to make to come out under budget and still have good data.  It’s what we do everyday.

I hope your day is going well.

 

Image courtesy of www.futureatlas.com/blog/

Share Your Thoughts...
Tell us what you're thinking...
and oh, if you want a pic to show with your comment, go get a gravatar for free here!
Comments are moderated and may not appear immediately...