Poynter's E-Media Tidbits mentioned a month ago how polls on news Web sites are inaccurate and are meant just for fun.
Yes, there's much that can be improved about them. Particularly on sites that have implemented mandatory registration.
The Chicago Tribune, for instance, requires readers to register an account in order to vote in polls (example poll). And it requires each registered user to have a valid e-mail address not used by any other user. But you can vote as many times as you want -- as long as you clear the appropriate cookie between votes.
The case appears to be the same on AJC.com, the Atlanta Journal-Constitution's Web site. (Disclaimer: I once worked there.) On a story with a related poll (example story), you can vote as many times as you want, as long as you clear cookies. And you have to be a registered user to be able to vote.
This is inefficient. Why not tie the two systems together, so that a registered user can only vote once?
Everyone would win. The quality of poll results would increase (although it would still be non-scientific), and registered users would have an incentive -- albeit a small one, but these things add up -- not to share their account information.
If you're going to require people to register to vote in a poll, you might as well tie their vote to their registration.
Posted by anonymous on November 19, 2004, at 7:47 a.m.:
I think you didn't think much about why they possibly didn't implement such a scheme. If you think about it, tying their vote to their registration means, somehow checking out if they have voted in that poll before. To be able to implement this, you can keep some piece of information in the user's account (session whatever) per poll, if you want to keep past polls open to those who didn't vote, or you can retstrict the lifetime of a poll and just have one field in user's session that will be reset everytime a new poll comes out. In the second case, it is easier to implement, provided that there is only one poll, or predictable number of polls per week (assuming all polls expire within a week). So it becomes a little complicated and when you look at what you get in the end, you may decide not to implement this costly feature at all, since the poll results will not mean anything in either way, i.e. just because you can restrict a user to vote once in a poll doesn't mean your poll results become meaningful. So I assume the people who think about these issues thought more in depth and decided that it doesn't worth to increase the load and complexity on the server for something that is not much important.
Imho, pollings are just a way to engage the user. The user will voice their opinion and if someone tries to distort the results, the other side will get angry and they will try harder and this will result in more number of page views which is a good thing for the news site.
So, there are more dimensions to the problem.
Posted by Adrian on November 19, 2004, at 8:58 a.m.:
Thanks for the comment, anonymous. I'd say I'm well aware of the technical/performance issues involved, having developed a number of registration systems, poll systems and other such CMSes for several news sites professionally. (Check out lawrence.com, for instance.)
Yes, limiting votes per user would require the site's database/storage system to keep track of votes for each user. But I would not call that a "costly feature." Hard-disk space is cheap, and doing the extra lookup could require as little as a single database lookup when a person votes. ("Has person X voted in poll Y?") Compared to the rest of the registration system, that's not a big deal.
Clearly, I don't know the technical details of the two examples I pointed out, so you're fair in bringing that up. But I just don't buy the performance excuse in this case.
Posted by Anonymous again ;) on November 20, 2004, at 12:24 a.m.:
Adrian, I didn't mean to say that it is a big deal, but in the end it is an extra cost both in performance and in effort to develop that system if you already don't have one. But more important than the cost, if you look at what you get when you implement it, I think you get nothing, because the poll results will not mean anything anyway, except if the polls are about your own site and your users satisfaction, in which case I agree with you, since the results matter. However, if it is a political question like who will you vote for, then who cares, because your results doesn't mean anything, unless as a site owner you want to know how will your audience vote. On the other side, news sites like controversy because that's what pulls audience to them. So if you can create a poll where users can vote endlessly, you might actually profit from that system.
Other than these, you are fundementally right though, because currently the quality of polls are very low, usually they don't mean much, but by implementing what you mentioned, people at least could learn something, even though that something is not everything.
Thanks for the discussion.
Posted by Zach on December 17, 2004, at 4:18 a.m.:
You could just for every registration, create a csv file in some directory (that nobody would ever try to open:) - and just have it preset up for like a thousand 0,0,0,0,0,0,0,0 - each one in succession is the next poll - 0,0,1,1,0,0,0,1,0,0,1,0,0,0,1,0 - and you know if they voted - for a few megs of space for tens of thousands of users - and its hardly a performance drag to just as they hit a poll - open up the csv that has the same filename as their registration number --- 84930.csv - and find out if the voted in poll 324 - heck you could even keep it as 0,0,1,4,3,0,0,0,4,0,3,2,5,2,5,3,0 - if you wanted to keep past specific votes.
That much work for a computer would bore an 8088 to sleep.
Comments have been turned off for this page.