One of the most frequent questions we get about customer surveys is, "What kind of response rate can I expect?"
This turns out to be a complicated question--starting with the issue of how you measure response rates.
In principle, response rate is the number of completed surveys divided by the number of people who were offered the survey. If you asked 100 people to take a survey and 25 actually took the survey, then your response rate is 25%.
In practice, when doing phone surveys after a customer service experience there are a some other issues to consider:
- If the customer doesn't answer the phone or hangs up before being transferred to the survey, does that count as "offering" the survey? This is the majority of survey attempts in most cases, so this makes a huge difference in measured response rate.
- If you call the customer but get a wrong number, does that count?
- If you decide not to count "no answers" as offered surveys, but the customer had agreed to take the survey in advance through an opt-in process, do you still count the "no answer" as not offered?
How you decide to measure response rate can change the reported response rate by as much as a factor of ten: a 2% response measured one way could be as high as a 20% response when measured another way.
Vendors, of course, like to use the most favorable measurement (which generally boils down to counting as few attempts as possible). This is how some companies can quote very high response rates for survey processes which don't actually manage to survey very many customers.
I prefer to use the simplest measurement possible: every attempt to reach a customer counts as an offered survey, whether that customer answered the phone or not, and whether or not we had the right phone number.
The advantages of this calculation are that it's hard to manipulate, it's easy for the client to understand, and it presents a realistic picture of how many surveys we can actually complete from a given customer base. It will also expose problems with the survey process (bad source data, for example) which might otherwise be hidden.