The Customer Service Survey
VocaLabs' weblog providing news and commentary on the challenges of providing good customer service.
I should have said "cranberry sauce"
Tuesday - October 11, 2005 11:50 AM in
Last night my medical provider finally tracked me down long enough to give me a survey about a recent visit. I like the care I'm getting, so I really wanted to be positive about this survey. Especially since it's not every day that I'm on the receiving end of a customer service survey.
It started out well. A polite woman called me up and explained that she wanted me to take a survey about my recent visit and it should take ten minutes. It started to go downhill when she told me that she would transfer me to a speech recognition system to conduct the actual survey. (Actually, I was amused by the prospect, but it wasn't a good sign. "We care enough about your opinion to try calling you ten times, but not enough to actually have a human being listen to what you have to say.")
It started out well. A polite woman called me up and explained that she wanted me to take a survey about my recent visit and it should take ten minutes. It started to go downhill when she told me that she would transfer me to a speech recognition system to conduct the actual survey. (Actually, I was amused by the prospect, but it wasn't a good sign. "We care enough about your opinion to try calling you ten times, but not enough to actually have a human being listen to what you have to say.")
The computer told me in a friendly, pre-recorded female voice that I'd be asked a series of questions and I should answer: excellent,good,fair,or poor.By the time the first question came around I'd already forgotten the last two options. No matter, my answer to the first few were "good." By the third question I'd forgotten whether the first option was excellentor very good. So here I am with exactly one safe response for any question: good. Then I get a question where the answer clearly isn't "good." What would you say?
I should have said "cranberry sauce." None of the expected responses sound like "cranberry sauce." It might have given me the list of options. But, silly me, I said something that sounded too much like one of the options. I think.
I said "help." That along with "main menu" and "operator" can usually get you out of any bind with an Interactive Voice Response (IVR) system. The system took that response and went on to the next question. Did it log an error? Did it think I said one of the options? If so, which one? I sure can't tell.
After half a dozen more questions on the same mystery scale, it switched to a different kind of question.
"Please respond with no definitely not no probably not yes definitely or yes mostly."
Written out on paper you can parse that a lot better than I could hearing it. Fortunately, when I answered "yes" to the first question it shot back with:
"Do you mean yes definitely or yes mostly?"
So it does respond to errors! At least sometimes. I still don't know what it did with "help." To be honest, the most annoying thing to me was that this is a Likert scale (an ordered range of options) but they initially presented them out of order. That sounds downright jarring to my ear, though for most people it's just mildly confusing in a way they can't quite put their finger on. All in all, though, this series of questions went well-- I knew I could give the right answer.
They saved the worst for last.
"Do you have any additional comments about your visit?"
[Pause. Half expecting a beep. Finally my chance to explain my only complaint.] "Yes. At the end of the visit I needed to schedule another appointment and--"
"Now we would like to ask you about this survey. Please answer yes or no. Did you have any problems with this survey?"
"Yes."
"Please give any comments about this survey."
I fast-talked as many comments as I could within the few seconds they gave me. I think I got about three or four in. I just hope they care enough to listen to the recording. But I don't think they'll figure out my answer to the "help" question. And they certainly won't get as nuanced a response as they expect from the section where I only knew how to answer "good."
Bottom line: if you care enough to mail me a written survey, then follow up with a dozen attempts to reach me between the hours when I get home and when I go to bed, even though I don't answer the phone during dinner, you should care enough to have a human being take my responses. Or at least not cut me off before I have a chance to give a response.
Until now, I'd been neutral on the machine-vs-human survey administrator question. The fact of the matter is that designing a good IVR is difficult, and it takes less human effort to personally ask the questions.
Posted by David Leppik
I should have said "cranberry sauce." None of the expected responses sound like "cranberry sauce." It might have given me the list of options. But, silly me, I said something that sounded too much like one of the options. I think.
I said "help." That along with "main menu" and "operator" can usually get you out of any bind with an Interactive Voice Response (IVR) system. The system took that response and went on to the next question. Did it log an error? Did it think I said one of the options? If so, which one? I sure can't tell.
After half a dozen more questions on the same mystery scale, it switched to a different kind of question.
"Please respond with no definitely not no probably not yes definitely or yes mostly."
Written out on paper you can parse that a lot better than I could hearing it. Fortunately, when I answered "yes" to the first question it shot back with:
"Do you mean yes definitely or yes mostly?"
So it does respond to errors! At least sometimes. I still don't know what it did with "help." To be honest, the most annoying thing to me was that this is a Likert scale (an ordered range of options) but they initially presented them out of order. That sounds downright jarring to my ear, though for most people it's just mildly confusing in a way they can't quite put their finger on. All in all, though, this series of questions went well-- I knew I could give the right answer.
They saved the worst for last.
"Do you have any additional comments about your visit?"
[Pause. Half expecting a beep. Finally my chance to explain my only complaint.] "Yes. At the end of the visit I needed to schedule another appointment and--"
"Now we would like to ask you about this survey. Please answer yes or no. Did you have any problems with this survey?"
"Yes."
"Please give any comments about this survey."
I fast-talked as many comments as I could within the few seconds they gave me. I think I got about three or four in. I just hope they care enough to listen to the recording. But I don't think they'll figure out my answer to the "help" question. And they certainly won't get as nuanced a response as they expect from the section where I only knew how to answer "good."
Bottom line: if you care enough to mail me a written survey, then follow up with a dozen attempts to reach me between the hours when I get home and when I go to bed, even though I don't answer the phone during dinner, you should care enough to have a human being take my responses. Or at least not cut me off before I have a chance to give a response.
Until now, I'd been neutral on the machine-vs-human survey administrator question. The fact of the matter is that designing a good IVR is difficult, and it takes less human effort to personally ask the questions.
Posted by David Leppik
Posted at 11:50 AM by | | | |