Son of survey

This comment in the bad survey design thread got me thinking further about where/when/what to do with surveys. It’s not my primary tool so some of these reflections take me a little longer than someone who makes their living as a quantitative researcher, for example.

A tiny new restaurant opened in our tiny town of Montara – the Montara Bistro. I dropped by yesterday to pick up a menu and saw that they are already looking for customer feedback.

not_a_survey.jpg

So to some folks, this is a survey. But it’s next to useless.

Why? Their questions are not too bad, but they are conversational questions, and should be presented that way. They are the basis of a conversation. Handing someone a sheet of paper (with no room to fill in any response) and asking them for essays is ineffective. It’s not fair. These are the questions they want answers to, but sometimes you have to ask a series of questions to get that information. And you can’t decide ahead of time which questions to ask. You have to ask a question, listen to the response, and then choose your next one. You can’t do that on a piece of paper. You need to have real people talking to each other and exploring the issues that way.

Not to mention that the restaurant has been open for a day or two, and there’s a presumption of an in-depth relationship that hasn’t really been built yet. What do I think of the Bistro Vision? Ummm, I don’t care.

I love what this artifact tells you about the company; that they really want to get a dialog going. They don’t have the tools in place to do it yet. Maybe it’s backed up by the way they interact with customers who come in; I don’t know. But this won’t work at all.

And I think this sort of inquiry is what a lot of design students are doing; identifying some open-ended (i.e., requires the respondent to write sentences) questions and sending them out by email. Some people will respond. Some may even write a lot. But you can’t follow up unless you send out another email. And then it’s just a conversation.

As with everything you “send out” who it gets sent to is a factor. Sending something to 3 friends is a very different approach than something that is quantitative in nature.

Look at this artifact from a recent project (created by our partners, not us):
survey_sample.JPG

This contained 31 questions, only a few open-ended ones. There’s randomization where needed (so you can filter out order-effects, where the first or last item might be picked more frequently in a list), and a large enough sample so that results can be processed to lead to conclusions – comparisons between different factors (this is the stats part I’ve been talking about).

Attitude toward technology meets Age
Purchase habits meets Region (with Age)
Stores shopped meets Region (with Age)
etc.

Tons of work and tons of math goes into creating tables (that then get interpreted) like
255b6076.jpg

As Paul Hogan (sorta) said “That’s not a survey, now that’s a survey!”

I hope this brings a bit more clarity to the discussion.

Series

About Steve