The Polls That Don't Suck
Single-question polls outperform multi-question ones four to one — and most publishers are still building the wrong kind. The design rules that actually move the numbers.
Polls have a perception problem. They're treated as the lazy cousin of the quiz — something you ship when you didn't have time to ship anything else. That instinct is wrong by a factor of about four.
A well-designed poll is the highest engagement-per-effort format in the entire interactive kit. You write one good question and the audience does the rest of the work for you.
The problem isn't that polls don't work. It's that most of the polls publishers ship are built like cheap quizzes — and they pay the cheap-quiz tax.
Single question or nothing
The first decision is also the most-broken one. Pollster muscle-memory pushes editors toward multi-question polls — they feel more substantial, more "researchy." Readers do not agree.
Vote completion by question count
n = 6,400 sessions · publishers across news, sports, lifestyle
This isn't a finding about attention spans. It's a finding about contracts. The reader scans a poll, decides "I can do this in five seconds," and votes. Add a second question and you've broken the contract. Even the readers who finish vote at half the rate they would have on the single-question version, and they share at less than a quarter.
The four polls that actually work
After enough polls in production, four archetypes consistently outperform — and they're for four different jobs.
Four poll archetypes, four jobs
What each format optimizes for and the placement that matches
- Vote rate62%
- Best placementInline, 2nd fold
Binary, polarizing, opinion-shaped.
- Vote rate54%
- Best placementEnd-of-article
“Which one are you?” lightweight.
- Vote rate49%
- Best placementHomepage / category
Tied to a future, datable event.
- Vote rate44%
- Best placementNewsletter top block
Real question + transparent results.
A few notes on each.
Hot-takes
The hot-take is binary, opinion-shaped, and asked from a position. Not "What did you think of the finale?" — "Was the finale ending earned, or was it a cop-out?" The framing forces a side. Hot-takes share because voting publicly identifies you with a tribe; sharing publicly recruits to it.
If your question can be answered "yes, it was fine" — it's not a hot-take, it's a comment-section prompt, and it will perform like one.
Identity polls
"Which one are you?" Lightweight, never more than four options, always with a recognizable archetype label. The vote isn't the goal — the result reveal is. A good identity poll is a pre-quiz — a 5-second commitment that warms the reader up for the longer interactive next to it.
The single most underused trick in this archetype: link each result option to a specific follow-up article. Recirculation jumps from ~22% to ~38% the moment you stop showing a generic "read more" block and start showing one curated link tuned to what they picked.
Forecast polls
Tied to a dated future event — a match, an episode, an election, an awards night. They work because they create a reason to come back. The reveal is the event itself. The vote is the bookmark.
The mistake editors make here is closing the poll too early. Leave it open until the moment of resolution; the late voters share at twice the rate.
Data asks
A real question to which the audience reasonably believes their answer matters. "Are you finishing the season, or did you bail mid-way?" "How many alarm clocks have you owned this year?" The trick is that the result page must look like data, not like a popularity contest — a clean horizontal bar chart with absolute numbers, no confetti.
A poll feels like a poll when the result looks like data. It feels like a quiz when the result looks like a verdict.
Designing the question
Most poll authoring mistakes happen here. Five rules.
Ship these patterns
- Asymmetric options. "Loved it / Hated it / Hadn't seen it" outperforms three symmetric options because it tells you something true about the audience.
- One emotional verb in the question. "Did the show earn its ending?" beats "Was the ending good?"
- Anchor the comparison. "Better than season three or worse?" pulls in casual readers; "Was it good?" doesn't.
- Show vote totals after voting, never before. The pre-vote count anchors and biases. The post-vote count rewards.
- Cap option text at six words. Long options re-frame the question and lose votes to the shortest option, regardless of meaning.
Recognize the dead formats
- "What do you think?" as the question. It's not a poll prompt, it's a sigh.
- More than four options. Anything past four collapses into "the top two and noise."
- An "Other" option. Looks fair, performs like a tax — siphons votes from the real top option without producing useful signal.
- Voting bars before you vote. Single biggest vote-rate killer in the format.
- Anonymous-looking results. "47% / 53%" tells the reader nothing about who they're voting with.
The result reveal
The post-vote screen is what makes a poll feel earned. Three jobs, in order.
A poll without a result-reveal isn't a poll. It's a button.
Where to ship it
Polls at the end of an article are almost always misplaced. By the time a reader scrolls to the end, the bottom 30% of your audience has already left and the top 30% is gone too. The middle is your voter base.
A pre-publish checklist
- One question. (If you wrote two, kill the worse one.)
- The question contains one emotional verb.
- Three or four options, asymmetric, capped at six words each.
- No "Other," no vote counts shown before voting.
- Result page shows absolute numbers, the tribe split, and one curated next link per option.
- Placement is 2nd-fold inline or newsletter-top — not end-of-article.
- Standalone share URL with its own OG card.
A poll is the cheapest engagement you'll ever ship. That's only true if you treat it like the format it is — not a quiz wearing a smaller suit.
If you want the engine the better-performing examples above run on, the free plan ships polls with all of the above wired in by default — single-question, post-vote results, standalone share URLs, OG cards, and a native embed that doesn't tax your Core Web Vitals. See the pillar quiz playbook for the longer-form companion piece.
Ship interactive content your readers actually finish.
Quizzes, polls, personality tests, and photo battles — embedded as a native web component, not an iframe. Free plan available.