Polls, not surveys
January 2021
Running complex, multi-question surveys and applying quantitative methods to the responses is a pseudoscientific waste of time, and you should never do it.
If you disagree, please skim this article and then get in touch so we can argue đ
That said, single-question surveys (a.k.a. âpollsâ) are a great way to uncover unknown unknowns on your website, and you should absolutely use them.
Hereâs my favorite poll question:
⌠and hereâs whatâs great about it.
Itâs quick
Youâre not asking visitors to invest an afternoon in giving you feedback. Thereâs no need for a progress bar or an introductory screen that sets expectations or explains in detail how the responses will be used.
Because of this, youâll get more responses - even from the type of people who âusually donât fill these things out.â
Itâs humble
The question assumes that the page can be improved, and confers expert status on the visitor.
One of the major weaknesses of surveys is that respondents tend to give the answers they think you want to hear; this framing precludes that.
Where to put it
This poll works best on high-exit pages. (In Google Analytics, Behavior > Site Content > Exit Pages.)
The question assumes the visitor has looked for what they need, and failed to find it. So itâs best to show it after a delay. Depending on the density of information on the page, anywhere from 15 to 45 seconds could make sense.
What to do with the responses
First of all, ignore the visitors who inevitably type âyour momâ or other unhelpful answers. It does happen đŹ
After a few weeks, read through the responses and ask a few simple questions:
- Are visitors missing content thatâs actually on the page? Revisit how itâs presented.
- Are visitors missing content thatâs not on the page, but should be? Add it.
- Is there a recurring theme here? Think about how you can better answer these questions - on this page and elsewhere.
This is a qualitative exercise. It broadens your perspective, deepens your visitor empathy, and can feed into your experiment ideas.
Youâll still need to measure or test any high risk changes, e.g. adding pricing where it was previously absent.
But if you come away with the poll knowing just one thing you wouldnât have found out on your own, consider it a success.