Two forms get called a presentation feedback form and they have almost nothing in common. One goes to three colleagues who sat through a rehearsal and are expected to say something useful before the talk happens. The other goes to sixty people who have just watched the talk and will give it ninety seconds of attention at most. Using the same form for both is the reason so much presentation feedback arrives as a row of fours and the word interesting.
The questions differ, the anonymity setting differs, the length differs, and what can be done with the answers differs. Getting those four decisions right takes less time than designing the questions, and it matters more.
Two forms, two jobs
| Rehearsal form | Audience form | |
|---|---|---|
| Who answers | Two to five colleagues | Everyone who attended |
| What it is for | Changing the talk before it is given | Deciding what to do next time, and answering what was left open |
| Best length | Four to six questions, mostly free text | Three to five questions, one screen |
| Anonymity | Named, because it needs a conversation | Anonymous, with an optional contact field |
| When | Immediately after the run through | Before the audience leaves the room or closes the tab |
| Main risk | Politeness | Blandness at volume |
The rehearsal form fails when reviewers are polite. Colleagues who like the speaker will praise the slides and mention one typo, because volunteering that the middle third does not work feels like an attack. The audience form fails differently: it collects plenty of answers, and they are all variations of good talk, because nothing in the form gave anyone a reason to be specific.
Both failures are caused by the questions, not by the people answering.
The rehearsal form: make politeness impossible
The trick with peer feedback is to ask questions that cannot be answered politely. A reviewer who is handed a scale will use the top of it. A reviewer who is asked to produce a specific artefact has to engage.
Four questions do most of the work.
Write in one sentence what this talk was arguing. This is the single most valuable question on any presentation feedback form. If three reviewers write three different sentences, the talk has no thesis yet, and no amount of delivery practice will fix that. It also cannot be answered with encouragement, which is the point.
Where did attention drift, and what was on screen at that moment. Reviewers can always answer this honestly, because drifting attention reads as a fact about them rather than a criticism of the speaker. The slide reference turns it into something the speaker can act on in ten minutes.
Which single section would be cut if the talk had to lose three minutes. Asking for a cut forces a ranking. Asking what could be improved does not, and produces a list of small polish items while the structural problem survives.
What question is a hostile listener going to ask, and is it answered in the talk. This surfaces the objection the speaker has been avoiding, which is usually the reason the middle section feels defensive.
Notice that none of these ask for a rating. Scores are the wrong instrument for a rehearsal, because there is nothing to compare the score against and nothing in the number tells the speaker what to change. Keep the rehearsal form to free text, keep it to a handful of questions, and keep the reviewers' names on it. Peer feedback works when the speaker can go back and ask which slide, and that is impossible if the form was anonymous.
One practical addition: a field for the timing each reviewer recorded. Three reviewers reporting that the talk ran to twenty six minutes against a twenty minute slot settles an argument that the speaker would otherwise have with themselves.
The audience form: three questions that beat fifteen
An audience will answer a form on a phone, standing up, while other people are leaving the room. That is the whole design constraint. Anything that does not survive ninety seconds on a phone should not be on the form.
| Question as usually written | What comes back | Question that produces something usable |
|---|---|---|
| Rate this presentation, 1 to 5 | 4.3, every time, from every speaker | What is one thing from this talk that will change what you do next week |
| Was the content relevant to you | Yes, from people who meant not really | What did you expect this talk to cover that it did not |
| Rate the speaker's delivery | 4, with no indication of what to change | Where did you lose the thread |
| Any other comments | Blank, or good talk | What question were you left with |
The last one earns its place twice. It tells the speaker which part of the argument did not land, and it produces a list of real questions that deserve answers, which is a different kind of output from a survey. Those answers are correspondence: each one belongs to a person, needs somebody to write back, and needs a record that it was dealt with. A form that keeps each response with an owner and a status handles that; a spreadsheet of scores does not, and this is where most conference feedback quietly stops.
Keep one scored question if a number is needed for comparison across talks or across years, and keep it identical every time. A single consistent item is worth more than six that change with every event, because the only useful thing about a score is the trend.
Delivery is the wrong thing to put on the form
Almost every presentation feedback form asks the audience to rate delivery, and almost none of them get anything back that a speaker can use. Delivery problems are physical and specific: speaking to the screen instead of the room, filling gaps with a repeated word, racing through the part that was least rehearsed, standing still for forty minutes, reading the slide aloud. An audience notices the effect and not the cause. What arrives on the form is a four, or the word engaging, or the word monotone, none of which says what to do differently.
Two instruments work better than a survey question. The first is a rehearsal reviewer given one job, such as counting filler words or noting each time the speaker turned to the screen for longer than a few seconds. A single named observation with a count attached changes behaviour; a scale does not. The second is a recording of the talk. Ten minutes of playback at double speed shows the speaker every habit the form was trying to describe, and it removes the argument, because nobody disputes a recording of themselves.
That leaves the audience form free to ask about the argument, which is the thing only the audience can report. Whether the point landed, where the thread was lost and what question remains are all invisible from the stage and obvious from a seat. Asking the audience about delivery spends one of three or four questions on information they cannot give reliably, and it crowds out the one question whose answer nobody else can supply.
Ask before the audience leaves
Presentation feedback decays faster than almost any other kind. The specific moment where the argument slipped is available for a few minutes and gone by the evening, replaced by a general impression.
So the link has to be in front of people while they are still in their seats. A short URL and a code on the final slide, left up during questions, collects several times what an email the next morning collects. Say out loud that the form is three questions long, because the audience is estimating cost before deciding, and their estimate is usually a fifteen question survey.
For an event with several sessions, there is a choice between one form per talk and one form with a session field. One form per talk gives each speaker a clean link and a clean set of responses, and it means duplicating a form for every session, which is fast if the tool can copy a form with its questions and its notification settings intact. One shared form is less work to set up and pushes a sorting job onto whoever reads the results, since every response now has to be filtered by session before it means anything. For more than about six sessions, per session links are usually less total work, and they also stop a respondent answering about the wrong talk. Event use cases tend to settle this the same way.
Whichever way it goes, decide in advance who receives the responses. At most events the organiser collects feedback and the speaker sees a summary weeks later, if at all. Feedback that reaches the speaker after the next version of the talk has already been given has no effect on anything.
Read it without being flattered or wounded
Sixty responses of good talk and two specific criticisms will feel like a success and contain two pieces of information. Both of those are in the criticisms.
Count themes rather than reading impressions. Three people writing that the section on pricing was confusing is a finding. One person writing that the whole approach is wrong is an opinion that may be right, and it should be weighed on its content rather than on how forcefully it was expressed. The loudest comment is not the most common one, and the free-text column is where that distinction gets lost.
Separate what the speaker controls from what the speaker does not. Comments about the room, the timetable, the projector and the coffee are real, and they belong to the organiser. Filing them under presentation feedback makes the speaker feel judged for things they cannot change, and buries the two comments that were about the talk.
Look for the same sentence twice. Feedback that repeats across different audiences is about the talk. Feedback that appears once, at one event, is often about that room on that day.
And record what was decided next to the comment that prompted it. Two words, such as cut this section or keep, disagreed, turn a set of reactions into a record. A year later the question is always what was changed after the last run, and nobody remembers.
The part that outlives the talk
Most presentation feedback is collected once, read once and lost, which is why the same problems show up in the same talk three events running.
Three things make it durable. First, every response keeps its own status, so a question that needed an answer can be distinguished from a comment that needed nothing. Second, replies are sent from the same place the response lives, so the next person to look can see that the attendee was answered rather than guessing. Third, what the team recorded afterwards, such as the decision or the session it belonged to, exports along with the answers instead of living in a separate spreadsheet that will not survive the next handover. Those three are the difference between a form builder and a tool with response management.
That is also the honest test for any tool being considered for this. Submit one response to a three question version of the real form, then try to assign it to a colleague, mark it as answered, write a reply from inside the tool, and find that reply a month later. Rebuilding the whole form during a trial tests the part that matters least.
What to change first
Split the rehearsal form from the audience form, and cut the audience version to three questions that fit one phone screen. Put the link on the final slide rather than in an email the next day. Then make sure the questions people were left with get an owner and a reply instead of a row in a spreadsheet, with Halict or any tool built the same way.
Q1. What questions should a presentation feedback form ask?
For a rehearsal, ask reviewers to write in one sentence what the talk was arguing, where their attention drifted and what was on screen then, which section they would cut to save three minutes, and what a hostile listener would ask. For a real audience, ask what they will do differently, where they lost the thread, and what question they were left with. Ratings are optional and should stay identical between events if they are used at all.
Q2. Should presentation feedback be anonymous?
For a real audience, yes, with an optional contact field so that anyone expecting an answer can get one. For a rehearsal, no. Peer feedback is only useful if the speaker can ask which slide or which minute, and an anonymous rehearsal form removes exactly that conversation.
Q3. How many questions can an audience form have?
Three to five, on one screen. The audience is answering on a phone while the room empties, and every extra question trades a few completed responses for a little more detail from the people who would have answered anyway. Saying out loud that the form is three questions long measurably helps, because most people assume a form is longer than it is.
Q4. Is a rating out of five worth collecting?
Only as a trend. A single score cannot say what to change, and almost every speaker at almost every event scores around four, so the number flatters without informing. If a score is collected, keep the wording and the scale identical across events, and read the direction of travel rather than the value.
Q5. What should happen to the questions attendees leave in the form?
Treat them as correspondence rather than survey data. Each one needs somebody named to answer it, a record that it was answered, and the answer kept alongside the original message. Scores can be summarised and filed; unanswered questions from named attendees are the part of a presentation feedback form that damages a reputation when it is ignored.