Are People Ready for Voice Surveys?

In our increasingly connected world, smart technology has woven its way into nearly every aspect of daily life. The convenience of interacting with technology through spoken commands has become commonplace: more than 6 out of 10 Americans have used a smart speaker or a voice-assistant. Voice recognition and transcription technology have now matured. They are ready to be used in market research. But are respondents prepared to use their voice in quantitative surveys?

Voice Surveys: A Paradigm Shift?

Imagine a survey where instead of typing responses, participants are asked to answer open-ended questions using their voice. This methodology taps into the naturalness of conversation, potentially reducing the friction often associated with typing on smaller devices. Advocates suggest that this approach could yield more authentic responses, capturing nuances that text alone might miss.

In theory, respondents like this option. IntelliSurvey’s latest research-on-research project surveyed a representative sample of the US population, revealing that 42% were strongly in favor of being offered an option to record their thoughts instead of writing.

Participants shared sentiments such as:

“[C]ould be easier and [faster] for someone who isn't as keyboard savvy” “I think it makes it more inclusive” “It allows for more expressive responses.”

Recording in browser/ app is easy enough. Transcription technologies are now fairly accurate, outperforming humans on most tasks. So, should we embrace this new method fully? What are we waiting for?

Current Adoption: A Critical Project Risk

While people appreciate having options, utilizing them when offered is another matter entirely. Participants in this study were asked to share their thoughts on public services in their area, offering them the choice to type or record their responses. The results were telling:

• Less than 7% tried clicking in the recording button area.
• Only 4% managed to record something.
• Only 2% recording something that sounded intentional.
• And a mere 1% recorded a satisfactory open-end answer.
• In comparison, 90% of those who typed their answers provided satisfactory open-ends.

This critically low adoption rate underscores significant barriers to the widespread use of voice surveys. The force of habit is the main reason respondent quote to explain their choice to write (37%). Some have a stated preference for writing (ability to edit, better answer accuracy, and a general dislike for voice-recording). Additionally, situational factors (inability to talk, background noise), technical challenges, and privacy concerns also emerged. As some respondents expressed (in writing):

“No thanks. I’m still writing” “I worry that recording instead of writing could open my up to [vulnerabilities] or important background [information] may also be recorded.” “It seems easier but I just don’t like to record” “I DO NOT HAVE A MICROPHONE OR CAMERA FOR MY HOME COMPUTER.” “I don't like the Record Option, I like to think my answers through and if something does not look correct I can go back and change it where in a recording I have to go back delete the entire recording and start over again it just takes too much time and I don't like my voice [anyway].”

Based on this research, IntelliSurvey estimates that requiring respondents to provide satisfactory voice-recorded answers could decrease survey completion rates by a factor of 5. This mandate could also introduce specifics biases, skewing the overall survey results. For distribution list surveys, this could evoke strong negative reactions among contacts. In panel surveys, a qualification rate divided by 5 would substantially increase project costs, and could jeopardize the entire project feasibility.

Future Adoption: Some Encouraging Signs

Voice-to-text technology is not new, but recent advances in AI have brought it to the forefront. As such, it is understandable that it will take time for the general public to become comfortable with it.

Encouragingly, when respondents were asked follow-up questions after discussing voice surveys, the rate of satisfactory spoken open-ends increased to 4%. While still a small percentage, this indicates potential growth as familiarity with the format increases.

Survey panels could foster the habit of voice-recording by introducing it during the initial profiling of their panelists, presenting voice as a normal option from the start.

Despite the challenges, there are encouraging signs. For the few respondents who chose to answer using their voices, their responses were notably longer — approximately 70% longer before cleaning and 40% longer after removing repetitions and filler words — than those who typed. This suggests that voice responses could provide richer, more nuanced insights.

Conclusion: Not Now but Keep an Eye on Voice-Recording

For now, market researchers should temper their expectations and continue to rely on traditional text-based open-end input. What we are waiting for is for respondents to be ready. At time of writing, the possibility of slightly richer content does not justify the high dropout rates risk.