Surveys (the no-AI engine) hide the conversational types (AI Question and Exploratory) and use structured types only.
Open response
AI Question
The signature Versive question: an open-ended question where the AI interviewer probes with intelligent follow-ups based on what the participant says.Exploratory
A goal-driven, multi-turn AI conversation. Instead of a fixed question, you give the AI a research goal and a turn budget (default 8 turns), and it explores the topic conversationally until the goal is covered or the budget runs out. Exploratory questions support the same duration tracker settings as AI Questions:
The timer resets for each new question or follow-up. It pauses after 30 seconds without an initial response and resumes when the participant types or records; stopping a recording also pauses it. Reaching the maximum shows a reminder to wrap up, without forcing the participant to advance. These targets apply to individual responses, not the entire exploratory conversation.
Targets must be non-negative whole seconds, and the minimum cannot exceed the maximum when both bounds of a response range are explicitly set. A single follow-up override is allowed even if it falls outside the automatic bound; in that case the timer may have no green interval.
Turning the tracker off in the editor clears its duration targets; disabling it through the assistant or MCP preserves them for reuse. Ask the assistant to clear a specific target to remove it, or clear a follow-up override to restore automatic timing. MCP callers can pass
clear_duration_fields to edit_question with a list containing min_duration, max_duration, follow_up_min_duration, or follow_up_max_duration; passing null for a duration argument keeps its stored value unchanged.
Open Text
Long-form free text (or voice) input. Supports optional instructions shown to participants, and an optional character limit — participants see a live counter and typing stops at the limit.Short Text
A brief, single-line text answer. Supports an optional character limit with a live counter.Message
Not a question — displays information, instructions, or media between questions. Settings: custom button text (default “Sounds good”) and an optional timer that participants must wait out before continuing.Choices & sorting
Multiple Choice
Single- or multi-select from a list of options.Matrix Selection
Participants pick one answer for each row, or several with Multi selection on: rows are the items, and the answer choices are shared by every row. Use it when the answers are labelled choices rather than a numeric scale (for that, use Matrix Rating Scale).
Results show the answer distribution for each row. With Multi selection, each row’s percentages are over the participants who answered that row, so they can add up to more than 100%. The CSV export has one column per row holding the chosen answer; with Multi selection it has one column per row and answer (the answer when picked, blank when not), plus an “Unmatched” column per row for answers renamed or deleted after they were picked. A later question can reference the answer for one row with
selectionForItem[index]; several answers are joined with ”, ” (see Referencing earlier answers). A question with Multi selection on can’t be a display logic source.
Card Sort
Participants sort cards into categories — ideal for information architecture research. Settings: cards, categories, randomize order, allow new categories (open sort), and allow new cards.Ranking
Rank a list of options by preference, via drag-and-drop or dropdown format. Supports randomized order.Matrix Ranking
Rank options across multiple criteria (e.g. rank features by importance and by satisfaction). Same format options as Ranking.Allocation
Distribute a fixed budget of points across options (default total: 100). Options can optionally be pulled from a previous question’s answers, and you can enforce the total as a hard limit. With the limit on, options left blank count as 0 (and are saved as 0) once the values add up to the total, so participants don’t have to type 0 into the rest. The other rules below still apply: a custom option needs a name and a value, and a failing check against an earlier answer still blocks. Turn on custom options to let participants add their own options, each with a name and a value; their values count in the total, and a name that is already in the list is refused. With the limit off, each option is a separate value, and an optional range (a Min and a Max) applies to every value, including the options participants add: for example 0 to 100 for a percentage per option, or -100 to 100 for growth that can also be a decline. An empty Min means 0, an empty Max means no maximum. Participants see the range under the options and cannot continue while a value is outside it. With the limit on, every value must be 0 or more; when the total must equal an earlier answer that is negative, every value must be between that total and 0 instead, and when the total must be at most such an answer, every value must be that total or more. The total, or each option, can also be checked against an earlier answer — see Checks against earlier answers.Matrix Allocation
A grid of values: rows are the things being split, columns are scenarios or time points (up to 8), and each column is filled in on its own. With Has allocation limit on (the default), every column must add up to the total allocation (default 100), and participants see a live total under each column. With the limit off, each cell is a separate value, an optional range (a Min and a Max, the Min may be negative) applies to every cell, and columns don’t need to add up. An empty Min means 0, an empty Max means no maximum. Participants type numbers with up to two decimal places. With the limit on, cells left blank count as 0 (and are saved as 0) once the column adds up to the total. With the limit off, every cell needs a value. Columns never affect each other. Click a cell in the N/A cells grid to mark it not applicable (for example, a product launching in 2027 has no 2026 value): participants see it greyed out as N/A and can’t enter a value, it doesn’t count toward the column’s total, and a column whose cells are all N/A shows only N/A, with no total to reach. Supports randomized row order; columns keep their order. Results show one Allocation-style breakdown per column, leaving N/A cells out of the averages; exports show N/A in those cells.Conjoint
Choice-based conjoint analysis (CBC): participants repeatedly choose between product concepts assembled from your attributes, revealing how much each attribute drives preference.
Results include attribute importance, level utilities, choice shares, and design diagnostics — all computed automatically in Insights.
Tree Test
Findability testing on a navigation tree. Participants complete tasks by locating the right node in your hierarchy.Ratings & feedback
Rating Scale
A Likert-style scale (default 1–5) with customizable labels for the low, middle, and high points.Star Rating
A configurable icon rating — stars, hearts, or paws — with an adjustable maximum (default 5).NPS
Standard Net Promoter Score (default 0–10) with customizable end labels (“Not at all likely” / “Extremely likely”).Matrix Rating Scale
Rate multiple items on the same scale. Supports randomized row order and an optional N/A column. An optional one selection per column constraint makes each scale value usable by at most one row — once a row picks a value, it’s greyed out for the others (N/A stays shared). Useful for forced-ranking grids. The toggle is only available when every row can still receive a value: at least as many scale points as rows, or the N/A column enabled.Other
Prototype Task
A Figma prototype navigation task. Participants attempt to complete a task in your interactive prototype while Versive tracks their path.
Task outcomes are tracked as not started, in progress, completed, or gave up.
Website Task
Send participants to a live website or prototype to complete a task.Paste the published prototype link — the
figma.site, vercel.app, lovable.app, bolt.host, replit.app, or framer.website URL — not the editor or chat share link. For Figma design prototypes (figma.com/proto/...), use the Prototype Task instead. See Prototype testing for the full setup guide.Website Link
Redirect the participant to an external URL (e.g. a partner survey or booking page).Media Upload
Participants upload images and/or videos. Configure allowed types, maximum number of files, and maximum file size.Number
Numeric input, with an optional min/max range validation. The number can also be checked against an earlier answer — see Checks against earlier answers.Date
Structured date entry with month/day selectors and a typed four-digit year. Full date is the default; authors can instead collect month & year or year only and choose MDY, DMY, or YMD display order. Optional earliest/latest limits restrict which dates participants can submit. Answers remain stored in sortable ISO partial-date format:YYYY-MM-DD for full dates, YYYY-MM for month and year, and YYYY for year only. Storage is independent of display order.
Calendly
External Calendly booking — share a scheduling link (e.g. for a follow-up session) inside the study.Referencing earlier answers
Questions can reference earlier answers with mentions — type@ in the question editor to pipe a previous answer into the question text. See Reusing earlier answers.
Checks against earlier answers
A Number or Allocation question can require its answer to compare with an earlier answer — for example, “patients who started treatment this year” must be at most “patients under your care”, or an age-group allocation must total exactly that number. Open the question card and turn on Checks against earlier answers, then pick the earlier question (a Number question, or an Allocation question and one of its options) and the comparison: less than, less than or equal to, equal to, greater than or equal to, or greater than.
A Number answer, and each Allocation option, can carry several checks; the Allocation total carries one. Several checks on one answer are joined with “and”: every check must pass. The builder does not stop you from writing checks that no answer can satisfy, such as “equal to Q1” together with “less than Q1”; a participant who meets such a pair is stuck on the question, so check each pair against the answers you expect.
Participants see the check as a hint (“Enter a number 78 or less”; for a total with an “equal to” or “less than or equal to” check, “78 remaining”) and cannot continue while a typed answer fails a check (an optional question can still be left blank). With the limit on, once the values add up to the total, an Allocation option left blank counts as 0 and is checked as 0. A check that fails prints one line under the answer, in the participant’s language, such as “The number must be at most 100.” or “The total must be at least 1200.”; an “equal to” or “less than or equal to” check on an Allocation total shows the remaining count instead. If the earlier question was skipped, or answered with something other than a number, the check does not apply — the card warns you when the earlier question is not required. When a participant goes back and changes an earlier answer that a check reads, the later answers stay; the later question is checked again against the new value when the participant reaches it, and it blocks until the answer fits.
Renaming an option keeps its checks; deleting one drops them. If the earlier question is deleted, changes type, or ends up after the question it checks (including through randomization), the card flags the check so you can fix it. To also show the earlier value in the question text, add a mention. Simulated interviews respect these checks too.