Running live Q&A and polls

A floor microphone reaches about four people per session, and it selects for confidence rather than for the quality of the question. That is the problem app-based Q&A solves, and it solves it through one feature: upvoting. Everything else is plumbing.

Done well, Q&A changes the character of a session. Done badly it produces an empty queue, a moderator improvising, and a speaker who never sees a question. The difference is mostly operational.

The moderator is the whole job

Q&A does not fail in the software. It fails at the front of the room. Somebody has to own it, and their job is specific:

• Open the queue at the start of the session, not at the end.
• Approve questions into the visible queue as they arrive.
• Merge duplicates, so one good question does not split six ways across the votes.
• Watch the sort order and notice what the room is converging on.
• At the five-minute mark, hand the speaker the top three.
• Read questions aloud to the room, so remote and hard-of-hearing attendees get them too.

That is a real job for one person per room. At a multi-track conference it is the most common under-staffing mistake, and the symptom is a queue nobody manages and a speaker who says “I think we’re out of time”.

Moderate before display

There are two models and one of them is better.

Approve into the visible queue. Questions arrive in a moderator view, get approved, and appear to the room. Attendees see a curated list; the moderator’s decisions are invisible.

Show everything, remove problems. Questions appear immediately and the moderator deletes what should not be there.

The second sounds more open and behaves worse. A visible unfiltered queue changes what attendees are willing to ask — the ones who most need the anonymity see everything published instantly and think better of it. It also forces the moderator to remove things in public, which is worse for everyone than never showing them.

event67 moderates before display: questions reach the organizer’s approval queue and are approved into the room. That choice is deliberate and it is the right default.

Upvoting, and why it matters

Sorting by votes changes what gets asked, not just what gets answered.

• The question the room actually has rises, rather than the question the most confident person has.
• Attendees who would never take a microphone participate, because voting costs nothing.
• The speaker gets a signal about what the room wanted, which is useful even for the questions there was no time for.
• Duplicate questions consolidate, if the moderator merges them.

One caution: votes reflect who is looking at their phone in minute twenty, not the whole room. Treat the order as strong guidance and reserve the moderator’s right to pull up question seven because it is better.

Anonymous questions

Allow them. With moderation before display, the risk is manageable, and the questions people will not put their name to are frequently the most valuable ones in the room — the ones about cost, about failure, about the thing everybody knows and nobody says.

If a specific session genuinely needs attributed questions, say so from the stage at the start rather than turning the option off silently.

Running polls

A poll is worth running when the result changes what happens next. That is the whole test.

Good polls

• “How many of you have shipped this in production?” — the speaker branches on the answer.
• “Which of these three should we cover in the remaining ten minutes?” — the room sets the agenda.
• “What is your biggest blocker here?” — the answer shapes the next five minutes.
• A prediction at the start, revisited at the end.

Bad polls

• Anything the speaker will read out and then continue past regardless.
• Demographics you already have from registration.
• Multiple-choice questions with a correct answer, unless the session is a quiz.
• More than two in a forty-minute session.

Response rates fall off a cliff after the second poll. Spend them.

The mechanics

Write the poll before the session. A poll composed live is a poll with a typo on a screen in front of four hundred people.

Keep options to four. Five on a phone screen means scrolling and a lower response rate.

Show the results. A poll whose results the room never sees is a survey, and attendees learn the difference within one session.

Close it. An open poll that keeps moving while the speaker is talking about the result is distracting.

Do not put a poll in the first two minutes. People are still arriving and settling. Minute five onward.

Session-by-session choices

Session type	Q&A	Polls
Keynote, 500+ people	Yes, moderated, last ten minutes	One, early, to wake the room
Breakout talk, 40 minutes	Yes, open the queue from the start	One, if it changes the talk
Panel	Yes — this is where upvoting matters most, because panels drift	No. The panel is the interaction
Workshop	No. The room is already talking	No
Lightning talks	Collect across the block, answer at the end	No
Closing session	Yes, and take the top-voted question of the whole event	One, on next year

The panel row is worth dwelling on. Panels drift more than any other format, and an upvoted question list is the only reliable instrument for pulling one back to what the audience came for.

Failure modes, and what to do

Nobody asks anything. Seed one. The chair or moderator asks the first question out loud, and the queue fills behind it. Silence is about permission, not interest.

One person asks six questions. The moderator merges or declines. This is why moderation before display is easier: nothing has to be removed in public.

A hostile question arrives. Approve it if it is a fair question asked sharply; decline it if it is an attack. The difference is whether an honest answer exists. If you decline, the asker sees nothing happen, which is the outcome you want.

Questions arrive after the session ends. Answer them in writing, in the app, that evening. Ten minutes of a speaker’s time, and it is the follow-up attendees remember.

The venue Wi-Fi is unreliable. Attendees will use cellular, so test on cellular. Tell speakers what to do if the queue stops updating: keep talking, take a hand.

The screen shows the queue and the room reads instead of listening. Do not put the live queue on the main screen during the talk. Poll results, yes. The queue, at the end.

Briefing the speakers

Q&A fails at the front of the room, so brief the front of the room. Three sentences in the speaker pack:

1. Questions come in through the app during your talk, and the moderator will hand you the top three at the five-minute mark.
2. Mention it in your first minute: “there’s a question queue in the app, use it as we go”.
3. Read the question aloud before answering, so the room and the recording both have it.

That last one is the most-skipped and the most-noticed. There is more in the speaker brief template.

What to keep afterwards

Every question asked at your conference is a record of what your audience actually wanted, and it is the best programme research you will get all year.

• Export the full question set, including the ones that were never answered.
• Sort by votes across the whole event, not per session.
• The top twenty questions are next year’s session list, and they are better than any call for papers because nobody was pitching.
• Send each speaker their own questions with the vote counts.

event67 archives every Q&A, poll and photo in an end-of-event report, which is where this comes from. It is the most under-used artefact most conferences produce.