Best practices for online polls

What do you think about online polls?

  • They’re a great way to drive traffic and build reader loyalty.
  • They are a misleading load of garbage.

    As a former statistics major in college, my reflex reaction is to choose option #2. But as an online editor, I think there are ways website publishers can use online polls responsibly… and effectively, to drive traffic and build loyalty.

    The first step toward doing that requires editors to understand the limits of online polls. And I’m talking about those widgets where anyone can click a radio button and hit submit, not controlled surveys open only to selected participants and run by experienced public opinion pros.

    The big problem with online widget polls is that they *are* open to anyone. That gives you a self-selected sample of respondents who most likely do not reflect society, or even your readership, as a whole.

    “Remember, the purpose of a poll is to draw conclusions about the population, not about the sample. In these pseudo-polls, there is no way to project the results to any larger group. Any similarity between the results of a pseudo-poll and a scientific survey is pure chance,” Sheldon R. Gawiser and G. Evans Witt wrote in their “20 Questions A Journalist Should Ask About Poll Results”, published by the National Council on Public Polls.

    If you are simply looking for little, easy-to-read nuggets of information to brighten a page, you don’t need to conduct a poll. There’s a wealth of research available, from the U.S. Census Bureau to that found in various publications of the American Statistical Association, which can give you those nuggets. (The ASA’s publications also provide great resources that can help reporters call out sources whose misinformation deserve refutation.)

    So, you aren’t generating anything with your online widget poll that’s going to give you accurate information about the public’s collective attitudes or behavior. So what’s left as a reason to run such a poll?

    Plenty.

    While widget polls won’t tell you anything useful about your readership’s *collective* behavior, they can engage your readers to share *individual* stories. They key is to stop thinking of these widget polls as little public opinion surveys and start thinking of them as… straw polls in a bar argument.

    In my experience, the best way to use widget surveys is as an introduction into a discussion on the topic raised by the poll. My personal favorite example, one that I’ve run on several websites, will make sense to almost all Americans who celebrate Thanksgiving: “Should cranberry sauce be whole berry or jelly-style?”

    It’s a silly question, sure. But it is one that most folks who have sat down to a Thanksgiving meal in the U.S. have personal experience with, and one that many of them are very willing to debate in public.

    Over the past months, my wife and I have been posting widget-poll “votes” to our websites every Friday afternoon. The questions have included asking theme park fans what they though should have been the new name for the Disney-MGM Studios theme park in Florida, and asking violinists to pick their favorite movement from Dvorak’s “New World Symphony”. We’ve found them a useful way to build traffic during what’s typically the slowest time of the week on the site.

    Widget polls tempt readers to click. They provide the easiest form of interactivity, easier than commenting, easier than chiming in on a discussion thread, easier than starting a blog. But once readers click, they’ve broken that invisible barrier between reader and website and become part of the community of the site, if only in a small way. That’s why widget polls can become an important tool for online community building. It’s a crass analogy, but widget polls are the “gateway drug” of online interactivity.

    As with any other content that a professional puts on his or her website, careful thought helps make these widget polls a traffic-building success. A poorly conceived or worded poll will not elicit clicks, responses or referrals from other URLs, making the poll a waste of everyone’s time — the publisher’s and the readers’.

    I groaned to myself this morning when I found the following poll on the website of the Pasadena (Calif.) Star-News: “Should California’s electoral votes be distributed by congressional district or the current winner-take-all system?” with the answer options “Yes” and “No.”

    Ugh. (To be fair to the Star-News, the answer options were clarified later that morning.)

    Anyone who wants to make their website polls a powerful feature on the site ought to spend a few moments reading the American Association for Public Opinion Research’s “Best Practices for Survey and Public Opinion Research”. The document is designed for public opinion pros conducting proper research, but at least two items can help Web editors with their widget polls:

  • “Take great care in matching question wording to the concepts being measured and the population studied.”
  • “Pretest questionnaires and procedures to identify problems prior to the survey.”

    (The second, for example, would have helped the Pasadena paper avoid its silly gaffe.)

    In addition, specifically for online widget polls, I would add these best practices:

  • Pick a topic relevant to your specific readership, upon which they will have some personal experience.
  • Ask a question that has a limited set of obvious potential responses. (Truly open-ended questions work better as discussion threads.)
  • Ask a question for which readers will want to explain their choice.
  • Invite and provide a way for readers to comment upon their choice.

    And, finally:

  • Take steps to prevent multiple votes from individual readers.

    As a Web editor of an interactive site, nothing makes your life hell more completely than having your site gain a reputation as an easy place where crackers and script kiddies can manipulate your content. If widget polls are the gateway drug to reader interactivity, easily gamed ones establish the gateway to reader-driven mischief and abuse. Setting a cookie after a vote, or blocking multiple votes from IP addresses, or both, can help you prevent ballot-stuffing.

    Many newspaper website content management systems include built-in online polling capabilities. Even if your system does not, you need not know how to code your own polling application. For the polls on my websites, I’ve been using the embedded widget from Twiigs.com, which I first encountered on the delightfully catty Go Fug Yourself.

    Twiigs hosts its polls on its website, and provides iframe coding that allows publishers to embed the poll on their own sites. That option makes it easy for an editor to direct readers to talk about their choice on the site, in a blog entry’s comments section for example.

    You might ask, if the point of this exercise is to have a discussion, why bother with the poll? Having the poll results on the page gives readers another, easy-to-see, point of entry into the discussion. One need not browse through multiple comments to get a sense of how the conversation is going. One can see that at a single glance with the poll results. And, as I wrote earlier, readers find the poll interface an easier and more welcoming interface than a comment button. (In my experience, we’ve gotten about one comment for every 15-20 votes cast in the poll.) Even if individual anonymity explains part of that wider appeal, I’ve found many readers on my websites make their first-ever comments to the site in response to a poll. Polls really can ease people across that threshold of interactivity.

    To be clear, I do not suggest reporting the percentage results of such online polls as news, or as representative of any group, even the readers of the site. At most, I might consider reporting the raw numbers of such polls (e.g. “250 readers of OJR.org reported that they’d flunked math in college, according to a poll on that site”), then pulling some interesting quotes from the comments. But the real value of online widget polls is not to create news for other websites. It is to give your readers one more engaging reason to spend their time on yours.

  • About Robert Niles

    Robert Niles is the former editor of OJR, and no longer associated with the site. You may find him now at http://www.sensibletalk.com.

    Comments

    1. I’m a big fan of giving readers plenty of choices in online polls. Those “yes / no” ones turn my stomach. Almost all the polls I write I add an “I don’t know” option for those people who don’t know but still want to participate. When the hometown baseball team was down 3 games to none in the World Series I put up a poll asking “What percent chance of winning the World Series do we have?” and gave poll options at 5% intervals. I’ve got no illusions about science here, I was just curious if people would vote on all 22 options (“I don’t know” included). People did — I’m going out on a limb here, but I think people like choice.

    2. It’s a very bad idea to ask polarizing questions in online polls. If you’re treating as entertainment, then you should stick to light topics.

      Our local weekly asked a question about a proposed park that was on the political faultline in our community. It’s common knowledge how to cheat on online polls. It was very clear from the flow of results that both sides were cheating and voting more than once because they knew that the results would show up in a tidy graph in Wednesday’s paper with all disclaimers in mousetype that no one would read. And sure enough, there it was.

      What a perfectly awful idea.

    3. I agree, Barry. Any question that you would ask in a “real” random-sample p.o. survey should not be asked on a Web widget poll.

      (But I would append a comment section to a report on the real poll, asking people to add their personal experiences to the report.)