Reading Survey Results: Summaries, Responses and Follow-Up

Overview

Collecting responses is the easy part. Making sense of them, and doing something useful with what people said, is where the work actually is. The Results section of a survey turns raw submissions into something you can read in ten minutes and act on the same afternoon.

It is built for the kinds of questions community organisations actually ask. Did the audience enjoy the show? Which night went best? Who has not yet returned their cast form? Which three people wrote something that needs a reply?

Who uses this: Organisation administrators holding the Surveys permission, found in the Organisation permission group.

At its simplest, results answer one question: What did people tell us, and what should we do about it?

Key capabilities:

How It Works

1. Read the shape of the answers

The Summary is where you land. Headline cards give you the numbers you would quote in a committee meeting, and below them every question is broken down in turn: how many answered, how the answers were distributed, and what people wrote.

This view is deliberately shallow. It tells you where to look, not what any one person thought.

2. Narrow to the group you care about

Filters for event date and ticket type appear where responses can be linked to an event. On the Summary they render within the summary itself, above the headline cards. On the Responses and Not responded views they render inside the report header. Choosing a filter narrows the Summary, the Responses list and any drill-down you open from a chart, not just the chart you happened to be looking at. This is how you compare Saturday night to Thursday matinee, or concessions to full price.

3. Click through to the people

Every chart segment and every written comment is clickable. Clicking opens the individual responses behind it, with previous and next buttons so you can read straight through a group rather than hunting for them one at a time.

4. Work the list

The Responses view gives you one row per response with a column per question, so you can sort, filter and search like any other report. From a row you can read the full submission or open that person's attendee record. If your organisation uses memberships, you can select rows and change membership in bulk.

5. Chase, print and export

The Not responded view lists everyone who was emailed and has not replied. The Summary can be printed. Both tables can be downloaded as a spreadsheet for records, funders or the committee.

The Three Result Views

Summary, Responses and Not responded are separate entries in the survey's section navigation, alongside Overview, Send and the question setup. They are not tabs inside a single results screen. You move between them by choosing the entry you want, the same way you move to any other part of the survey.

Each entry carries its own summary line in the navigation, so the response count and the date of the most recent response are visible without opening anything.

Why do forms open on the Responses table? A form collects required details from named people, such as a cast list or a volunteer sign-up. You almost never want an average, you want the list. Surveys and questionnaires are the other way round, so they open on the Summary.

Why is Not responded restricted to identified surveys? Listing who has not responded would reveal who has, which is exactly what an anonymous or confidential survey is meant to keep out of your hands. Seaty refuses to produce the list rather than undermining that by the back door. See Anonymity and data protection.

Headline Figures

Up to four cards sit at the top of the Summary.

Responses is always shown: the number of responses currently being displayed, after any filter.

NPS appears when the survey has a Recommendation question that has been answered. The card shows the score, with Promoters and Detractors underneath. The card is green when the score is zero or above and red when it is negative.

Average rating appears when the survey has a Rating question that has been answered, shown as the average out of the scale maximum, for example 4.2 / 5. If the survey has more than one rating question, the card is labelled with a shortened version of the question text instead of the generic "Average rating", so you can tell which number you are looking at.

Email response rate appears when the survey has been emailed and no filter is active. It is the number of invited people who actually replied, divided by the number of people emailed. Underneath it breaks down into "Replied to invitation", "Emailed", and, where relevant, "Link or QR" for responses that came from outside the invitation list.

Why can the response rate never exceed 100%? It only counts responses that can be matched back to someone you emailed. People who found the survey through a shared link or QR code are counted separately, under "Link or QR", rather than inflating the rate.

Why does the rate disappear when I filter? A filter changes which responses are counted but not who was invited, so the fraction would be misleading. It is hidden rather than shown wrongly.

Understanding the Scores

Recommendation (NPS)

A Recommendation question asks people to score 0 to 10, anchored at "Not at all likely" and "Extremely likely". The results screen calls the resulting metric NPS.

Answers are grouped into three bands:

The score is the percentage of promoters minus the percentage of detractors, so it runs from -100 to +100. A score of 0 does not mean nobody answered, it means promoters and detractors cancelled each other out.

Passives do affect the score. They are not added to either side of the subtraction, but both percentages are worked out against everyone who answered the question, passives included. Adding a passive therefore dilutes both figures and moves the score towards zero. It is a common misreading that passives are ignored entirely.

The per-question block shows the score, the three-part split, and a column chart of every score from 0 to 10 so you can see where people actually sat rather than only the summarised bands.

Why show the spread as well as the score? Two very different audiences can produce the same score. A crowd of straight 8s and a crowd split between 10s and 3s are not the same problem, and only the chart tells you which you have.

Ratings

A Rating question is a scale from 1 up to the maximum the organiser chose, anchored at "Poor" and "Excellent". The results show the average as a headline (for example 4.2 / 5) with a column chart of the distribution beneath.

Read the distribution before you quote the average. An average of 3 out of 5 produced entirely by people choosing 3 is a lukewarm audience. The same average produced by half 5s and half 1s is two audiences, and the interesting work is finding out which is which. Filtering by event date is usually the quickest way to find out.

How percentages are calculated

This is worth knowing before you quote a figure.

Why the difference? An unticked confirmation records nothing at all, so there is no such thing as a "skipped" confirmation to exclude. Counting against everyone who reached it is the only way "not confirmed" can be a meaningful figure.

Why "reached" rather than "total"? If a survey branches, a follow-up question that only some answers lead to was never put in front of the others. Counting a confirmation against everyone who responded would make it look as though half of them declined it, when in fact they never saw it. Reached counts only the people whose path actually took them there.

Each question header also states it plainly, in the form "12 of 15 answered, 3 skipped". The verb changes with the question type, so a confirmation reads "confirmed", and a document reads "agreed", "acknowledged", "read" or "opened" depending on what it asked for.

Charts and Bars

A Charts and Bars toggle sits on the Summary toolbar and switches the whole page between the two. Charts is the default.

When a survey has more than five questions, a jump list appears above the breakdowns so you can go straight to the one you want rather than scrolling.

Responses over time

Once responses span at least two days, a daily bar chart headed "Responses over time" appears. Quiet days are plotted as zero rather than skipped, so the gaps are visible. Dashed markers annotate the days an email went out, covering both invitations and reminders. Each marker names which it was and how many people it reached, in the form "Invitation sent to 240" or "Reminder sent to 88", so a spike can be traced to the send that caused it.

Reading Written Answers

Free text answers are the most useful and the most awkward part of any survey. They are handled separately from the charts.

Progressive disclosure. Six answers are shown at first, with a "Show all N answers" button to expand and "Show fewer answers" to collapse again.

Search and sort. Once a question has more than four answers, a "Search answers" box appears alongside Most recent and Longest sort buttons. Searching shows a live count, in the form "4 of 37 answers mention "parking"." If nothing matches, it says so rather than showing an empty list.

Why sort by longest? People who write at length usually have something specific to say, good or bad. It is a reliable shortcut to the comments that need a reply.

Every comment is clickable. Clicking one opens that person's full submission, and the list you were looking at, filtered and sorted as you left it, becomes the previous and next queue. Search for "parking", then read every one of those people in turn without going back and forth.

A cap you should know about. The Summary shows at most the 500 most recent answers for a question. When there are more, a notice appears above the answers saying so, and it also warns that search only covers those 500. To work through everything, download the Responses list instead.

Summarising Written Answers With Seaty AI

Two hundred written answers is more than most committees ever read. Seaty AI can group the answers to a single written question into themes, so you can see what people actually said without reading every line. It never runs on its own: nothing is sent anywhere until you press the button.

Where it appears. On each written-answer question in the Summary, once that question has at least ten answers. Below ten, Seaty says so and offers nothing, because a summary of six answers is worse than reading the six answers. It is not offered on forms.

What you get. A short summary of what people said, the themes that recur with a count of how many answers sit in each, a positive or negative reading for each theme, and a few suggested next steps drawn from the answers. Counts are worked out by Seaty from the answers themselves, not written by the AI, because counting is the one thing these models are unreliable at and the number is what ends up in your committee minutes.

The answers stay put. The summary sits above the answers, never instead of them. Every original comment is still there to read, search and click into.

The first time you use it

The first time anyone in your organisation presses the button, Seaty explains what is about to happen and asks you to confirm. That confirmation is the decision to use the feature: there is no setting to switch on beforehand.

What it tells you is worth reading properly, because you are the data controller for the answers your respondents gave you. The written answers are sent to OpenAI, which processes them in the United States. Only the answer text is sent, at most 500 answers and 500 characters of each. Names, email addresses and everything else about the respondent stay in Seaty and are never included. Seaty records who confirmed and when.

Confidential and anonymous surveys

On a confidential or anonymous survey the summary will never quote anyone word for word. It describes what people said in its own words instead, and Seaty checks the result and strips anything that reproduces an answer too closely before you see it.

This matters more than it sounds. A distinctive comment can identify someone even when no name was ever recorded, in the "the one who mentioned using a wheelchair" sense. On an identified survey, where respondents already know their name is attached, quotes are allowed.

Being honest about the limits: that check catches copying, not a paraphrase that happens to describe one recognisable person. The ten-answer floor and the paraphrase rule are the real protections, and on a small survey you should still read the result before you circulate it.

Cost, caching and refreshing

Running an analysis costs Seaty money, so the result is stored and repeat views are free. Reopening the Summary shows the analysis you already have without sending anything anywhere.

Refresh appears only when it is worth it. More answers arrive, and at some point the summary is out of date. Seaty offers Refresh once enough have arrived to change the picture: at least ten new answers, and at least a fifth more than were analysed last time. Below that it tells you where you stand rather than greying out a button with no explanation, for example "12 new answers since, 20 more needed to re-analyse". Thirty analysed answers need ten new ones; five hundred need a hundred.

A daily limit. Each organisation can run a limited number of analyses per day across all its surveys. It is set high enough that ordinary use never meets it.

When you cannot run it

While a filter or a search is active. The analysis always covers every answer to the question, not the ones currently on screen, so offering it mid-filter would invite you to read it as a summary of the filtered group. Clear the filter or search to run it. If an analysis already exists, it still shows, with a line reminding you that it covers everything rather than your current view.

After deleting a response or erasing a respondent. The stored analysis for that survey is deleted at the same time, because someone who asked to be erased should not survive inside a cached summary. The next run starts fresh. Archiving a survey does not delete anything: archive and restore, and the analysis is still there.

Treat it as a starting point

The panel says so, and it is worth repeating: the themes are inferred by AI. They are a way into the answers, not a substitute for them. Check a theme against the answers underneath before you quote it in a report, an application or a meeting. The counts are reliable because Seaty calculates them; the wording around them is a machine's reading of what people meant.

Drilling Into a Chart

Every segment is clickable. Doughnut slices, bar rows and histogram columns all open the responses behind them, with previous and next buttons to read through the group. Segments with a count of zero do nothing.

Drill-down respects the event date and ticket type filters you have set, so clicking the "1" column while filtered to Saturday shows the people who scored 1 on Saturday, not everyone who ever scored 1.

The "not confirmed" and "not agreed" segments work too. They are calculated by absence, so clicking one gives you the people who did not tick or sign, which is usually exactly the list you were after.

The Responses Table

The Responses view uses the same table Seaty uses for its other reports, so it behaves the way you would expect from Reporting: multi-column sorting, per-column filters, global search, showing and hiding columns (remembered per survey), a pinned first column, summary cards recalculated from the filtered rows, and a spreadsheet download.

Columns

Why are timestamps labelled UTC rather than converted? They are shown and labelled in UTC consistently, so a timestamp means the same thing in the table, in an export and to anyone else looking at the same file. Mixed local times across a shared spreadsheet cause more confusion than they solve.

Answers render sensibly per type. Single choice shows the chosen option, multiple choice shows the options separated by commas, rating and recommendation show the number as a real numeric column so it sorts and averages, free text shows the text, a confirmation shows "Yes" or nothing, and a document shows "Agreed", "Acknowledged", "Read" or "Opened".

Free text columns cannot be filtered. The filter control is deliberately hidden on them, because a filter list of two hundred whole paragraphs is unusable. Use the global search, or the search on the Summary, instead.

The summary cards above the table

The table carries its own cards, recalculated live from whatever rows the filters have left: Responses always, Average rating where the survey has one, and With email on identified surveys.

Row buttons

At the start of each row are up to two buttons. These buttons are the only way in. Clicking the body of a row does nothing, so use the buttons rather than the row itself.

Inside a single response

The response view shows every answer that person gave, with documents rendered read-only exactly as that person saw them. From here you can move through the queue, and delete the response if it needs to be removed. Deletion removes the response and its answers permanently, and it is recorded in the survey's audit trail.

On a confidential survey the response opens with a note saying the respondent's identity is hidden. A Reveal identity button sits beneath it only where a name or an email address was actually recorded against that response. Where neither was captured there is nothing to reveal, so the note appears on its own with no button. Revealing is recorded in the audit trail before the name is shown to you, and the screen tells you so. Read Anonymity and data protection before you use it.

Bulk Actions and Membership

Tick boxes sit at the start of every row on the Responses table. Select some rows and a bar appears with the actions you can apply to all of them at once.

Delete responses is always available, on every type of survey. It is marked as destructive, and the confirmation tells you how many responses it will remove. Deleting is permanent: the responses and every answer in them go, the scores and charts are recalculated without them, and any saved Seaty AI summary of the written answers is cleared so it cannot describe answers that no longer exist. Each deletion is recorded against the survey, so there is a record of what was removed and by whom.

Membership actions appear as well when your organisation uses memberships and the survey is identified. Both conditions are needed: on a confidential or anonymous survey Seaty will not hand you a list of names to act on, and without memberships enabled there is nothing to set.

Two membership actions are available:

Rows without an email address are dropped before either action runs. If none of the rows you selected have an email address, the dialog says so and there is nothing to apply.

Once applied, the table refreshes so the Membership column reflects the change. For more on member types, see Managing Members.

Why is this useful? A volunteer sign-up form or a membership renewal form produces exactly the list you need to update your membership from. Rather than copying names across by hand, you filter the table to the people who said yes and set them all in one go.

Chasing People Who Have Not Responded

The Not responded view lists one row per invited email address that has not responded, ordered by when each address was most recently invited, earliest first. The people you have left alone the longest are therefore at the top. You can re-sort by any column once the list is open.

A summary card shows the total not responded.

Someone counts as having responded if either their personal invitation link or their email address matches a completed response, which reduces the chance of chasing people who did in fact reply. It cannot be perfect. Somebody who ignored their invitation and instead used a shared link or QR code cannot be matched, so they may appear on this list even though they responded. The view says as much above the table.

To chase, send again from the Send screen with the "Only send to people who have not responded yet" option, or download the list and contact them another way. See Sending and sharing and Mail.

The event date and ticket type dropdowns have no effect on this list. They still appear in the header, and you can still change them, but the list you get back is the same either way. Someone who has not responded has no response, and therefore no event date or ticket type to filter on. If you set a filter here and the list does not change, that is why.

Filtering by Event Date and Ticket Type

Where responses can be linked to an event, two dropdowns appear: Event date and Ticket type. Each option shows its response count in brackets, so you can see at a glance whether a segment is big enough to be worth reading. The first options are "All dates" and "All ticket types", and a Clear filters button appears once anything is selected.

On the Summary these dropdowns render inside the summary itself, above the headline cards. On the Responses and Not responded views they render inside the report header. They apply to the Summary, the Responses table and any drill-down you open from a chart. They do not affect the Not responded list, as described above.

Event date labels always include the full date with the year, and are prefixed with the event name when the responses span more than one event. That matters because a survey belongs to your organisation rather than to a single event, so a form you reuse every season will naturally accumulate dates from several productions.

Ticket type labels are prefixed with the seat category where there is one, and are given extra detail when two categories would otherwise look identical.

If a filter matches nothing on an identified survey, the page says "No responses match this filter." with a button to clear it. That is a different message from the first-run empty state, which prompts you to share the link, QR code or email to start collecting.

On an anonymous or confidential survey you get the small-numbers protection message instead, because a filter matching nothing is below the five-response floor described below. It tells you how many responses the filter matched and invites you to clear it.

Small Sample Sizes: An Honest Word

This is the part most survey tools skip, and it is the part most likely to cause you trouble.

Below five responses, non-identified surveys show nothing. If a survey is anonymous or confidential and there are fewer than five responses, Seaty replaces the whole breakdown with a protection message instead of showing the numbers. This applies to the unfiltered view you land on first, not only when you have filtered. The same floor applies to the Responses table, so filtering cannot be used to get around it.

Why? With three responses, the numbers themselves identify people. If two out of three said they were unhappy and you know who the three were, you have effectively read a confidential answer. The floor is there so that a small anonymous or confidential survey does not become identifying by arithmetic.

Identified surveys are not suppressed, because respondents were told their answers would be attributed to them.

A single branched question can be hidden on its own, too. On an anonymous or confidential survey with pathways, a question that fewer than five people reached is hidden by itself, with the rest of the results shown as normal. The reason is the same one: a follow-up that only two people were led to would put those two people's answers on screen in isolation. The question still appears in the breakdown, but its numbers and any written answers are replaced with a short note in place of them.

If something abusive comes in below the floor, you can still remove it. The protection would otherwise trap you: with three responses to an anonymous survey, one of them defamatory, every route to that answer is closed by the very rules protecting the other two. So the protection message offers Review answers for moderation. It opens the written answers on their own, with a Delete button against each, and nothing else: no names, no email addresses, no ratings or choices, on any type of survey.

Seaty asks you to confirm first, and the confirmation is blunt about the trade. With this few responses, reading the answers may let you work out who wrote what, which is exactly what the floor prevents. It is there for acting on a specific answer somebody has complained about, not for having a look. Seaty records that you opened it, and when. If you then delete something, that is recorded separately, so an abuse case leaves a clear trail of who read the answers and what they removed.

This is not a way to see confidential results early. It shows written answers only, it is audited, and it does not reveal who wrote anything. Revealing a confidential respondent's identity remains a separate action with its own record against the individual response.

Above five, use judgement. The floor stops a specific disclosure risk. It says nothing about whether a number is worth acting on.

If in doubt, quote counts rather than percentages. "Nine of twelve people who answered said yes" is honest. "75% said yes" invites the reader to imagine a much larger survey than you ran.

Printing and Exporting

Printing the summary

A Print button sits on the Summary toolbar. Printing does three things for you automatically:

The printed page carries a header with your organisation name, the survey title, the response and invited counts, and which filters were active, labelled "All responses" when none were. That means a printout handed round a committee meeting states its own scope rather than being mistaken for the whole picture.

Exporting to a spreadsheet

Spreadsheet download comes from the Responses view or the Not responded view, not from the Summary. The file name is built from the survey title, the title of the report you downloaded, and the date you downloaded it. A survey belongs to your organisation rather than to a single event, so no event name appears in it. Expect something along the lines of "Cast Sign-Up_Cast Sign-Up responses_2026-07-19.xlsx": the survey title reads twice, once on its own and once as part of the report title. A file downloaded twice on different days is therefore easy to tell apart in a folder.

Because the anonymity rules are applied when the data is produced rather than by hiding columns on screen, an export carries the same restrictions as the screen does. A confidential survey exports without names and email addresses.

Common Questions

Reading the numbers

Why does my Summary show a protection message instead of results? The survey is anonymous or confidential and has fewer than five responses. Collect more responses, or clear any filter you have applied. Identified surveys are never suppressed.

Why is just one question hidden when the rest of the survey shows? On an anonymous or confidential survey that branches, a question fewer than five people reached is hidden on its own to protect that small group, while the questions everyone saw are shown as normal.

Why do two questions with the same number of answers show different percentages? Choice, rating and recommendation percentages are calculated from the people who answered that question. Confirmation and document percentages are calculated from everyone who reached that question, which on a survey without pathways is everyone who responded.

My average rating card does not say "Average rating". When a survey has more than one rating question, the card is labelled with a shortened version of the question text instead, so you know which rating the number refers to.

Why has my response rate card disappeared? It is only shown when the survey has been emailed and no filter is active.

What does a negative NPS mean? More people scored 0 to 6 than scored 9 or 10. Look at the histogram and read the written answers before drawing conclusions, particularly on a small survey.

Working with responses

Can I read all the responses from one night in a row? Yes. Set the Event date filter, then click any chart segment or open a row on the Responses table. The previous and next buttons walk through that filtered group.

Can I search written answers? Yes, per question on the Summary, once a question has more than four answers. Only the answers currently loaded are searched, which is at most the 500 most recent. Download the Responses list to work through everything.

Why can I not filter a free text column on the table? Filter controls are deliberately hidden on free text columns, because a list of whole paragraphs as filter options is unusable. Use the global search instead.

Can I edit somebody's answer? No. Responses cannot be edited after submission, by you or by the respondent. The only administrative changes are deleting a response or erasing a respondent's data.

Can I delete a single response? Yes, from within that response. It permanently removes the response and its answers, and the deletion is recorded in the survey's audit trail.

Identity and privacy

Why can I not see names on my survey? The survey is confidential or anonymous. Confidential responses can have their identity revealed one at a time, and each reveal is recorded. Anonymous responses carry no name or email address for the results screens to show, and there is no reveal button on them. See Anonymity and data protection.

Why is there no Not responded entry in the navigation? Either the survey has not been emailed to anyone, or it is not an identified survey. Producing the list on an anonymous or confidential survey would reveal who had responded.

Does exporting bypass any of this? No. The restrictions are applied when the data is produced, so the spreadsheet and the printout carry exactly the same columns as the screen.

Does the AI summary see who wrote each answer? No. Only the answer text is sent, never names, email addresses or anything else identifying. On confidential and anonymous surveys the summary is also written in paraphrase, and Seaty strips anything that quotes an answer word for word before you see it.

Someone has asked me to delete their answers. Does that cover the AI summary? Yes. Deleting a response or erasing a respondent also deletes the stored analysis for that survey, so their words cannot survive inside a cached summary. One honest limit: text already sent to OpenAI can persist in its own abuse-monitoring logs for up to 30 days, and an erasure inside that window cannot recall it. Nothing identifying was ever sent with it.

Seaty AI summaries

Why is there no Analyse button on this question? It only appears on written-answer questions with at least ten answers, and never on forms. It is also hidden while a filter or search is active, because the analysis always covers every answer rather than the ones on screen.

Why can I not refresh it? Not enough new answers have arrived to change the picture. Seaty shows you exactly how many more are needed rather than leaving a dead button.

Are the theme counts accurate? Yes. Seaty calculates them from the answers rather than asking the AI to count, because counting is what these models get wrong most often. The wording around the numbers is the AI's reading and should be checked against the answers before you quote it.

Memberships and follow-up

Why can I not select rows for bulk actions? Bulk membership actions require both that your organisation uses memberships and that the survey is identified.

What happens to selected rows without an email address? They are dropped before the action runs. If none of your selected rows have one, the dialog tells you and nothing is applied.

Someone appears on the Not responded list but I know they replied. They almost certainly used the shared link or QR code rather than their personal invitation link, so their response could not be matched to their invitation. The view warns about this above the table.

Printing and exporting

Will my charts print? Printing automatically switches to bars, because charts do not print reliably. Everything else prints as shown.

Will collapsed comments be left out of a printout? No. Printing expands every collapsed list of written answers first, then restores your view afterwards.

Does the printout show which filter was applied? Yes. The print header states the active filters, or "All responses" when none were set.