Add survey questions to your interview

A survey step collects a structured answer mid-interview. The AI moderator pauses, the participant taps their answer in the call interface, and the AI continues. Survey responses always show on the participant's screen, and the interview advances only when the participant answers. The moderator can't answer for them or skip past on its own.

Steps

1. Add a survey step

   In Build, click + Add element and choose one from the Survey group:

   • Multiple choice: pick one option (or several with multi-select).
   • Ranking: drag options into a preferred order.
   • Open text: a written answer (typed, not spoken).
   • Slider: pick a number on a sliding scale.
   • Rating scale: Likert, Faces, or Stars.
   • Media upload: upload a photo, video, or document as the answer.
   • Matrix: pick one option for each row of a grid.
   • NPS: rate how likely they are to recommend something, from 0 to 10.
2. Write the prompt

   Type the question. Keep it short: survey prompts work better as one clean sentence than as a paragraph.

3. Configure the type-specific settings

   Each survey type has its own rules and settings: see below.

4. Decide whether to ask a follow-up

   Most survey steps have an Ask follow-up questions toggle. Turn it on if you want the moderator to ask why they chose what they did: "why a 3 and not a 5?" produces the richest data.

Survey responses show up in two places:

• Inline in each interview's transcript, alongside the spoken responses.
• In Results → Analysis, grouped under each question (distribution charts sit above the qualitative themes).

Per-type rules

Multiple choice

• Add each option as its own row.
• Multi-select lets participants pick more than one option, with optional minimum and maximum selections.
• Other option adds a free-text "Other" choice.
• Exclusive options (multi-select only) adds an Exclusive section under the options. Drag an option such as "None of the above" into it: when a participant picks it, their other picks are cleared, and picking any other option clears it. An exclusive pick on its own completes the question, even with a minimum set. With a minimum of 2 or more, the hint participants see names the exclusive options too, for example "Choose at least 2, or 'None of the above'". Turning off Exclusive options or Multiple selection moves those options back into the main list.
• Randomize shuffles option order per participant. Exclusive options always stay at the bottom, in the order you set.

Ranking

• Needs at least 2 options.
• Randomize shuffles the starting order. The participant drags options into their preferred order.

Open text

• Participants type their answer into a text box, and Continue becomes available once they type something.
• Optional Character limit. With a limit set, the box stops accepting text at the limit and shows a running count, such as 42/200. Leave it empty for no limit.
• In analysis, the Responses card lists every answer with the participant's name. Search the answers or participant names, sort by participant, sentiment, date, or length, and see the response count and average word count at the top.

Slider

A slider starts with a range of 1 to 10 in steps of 1. Pick a Scale to set a common range in one click, or type your own minimum and maximum.

• Scale presets: Percentage (0 to 100), 0 to 10, 1 to 10, and 1 to 5. Each fills in the minimum, maximum, and a step of 1.
• Custom range: edit minimum or maximum directly. The Scale shows the preset your range matches, or Custom when it matches none.
• Step: choose from steps that divide the range evenly, so the participant can always reach the maximum. For steps of 10 on a 0 to 100 scale, keep the minimum at 0, since a 1 to 100 range doesn't divide by 10.
• The minimum must be below the maximum.

Rating scale

• Exactly 4, 5, or 7 points. No other point counts are supported.
• Three styles:
  • Likert: labeled agreement scale.
  • Faces: emoji faces.
  • Stars: star icons.
• Every point starts with a label filled in for you: Strongly disagree to Strongly agree for Likert, Very bad to Very good for Faces, and 1 star to 5 stars for Stars.
• On Likert, you can rewrite any point's label. If you clear a label, the original comes back when you click away, so no point is ever left blank.
• Changing the number of points updates the labels you haven't touched and keeps the ones you rewrote. Changing the style resets every label to that style's set.
• Faces and Stars labels can't be changed. If you need a fully custom set of options, use Multiple choice instead.

Media upload

Participants answer by uploading one file from their device. Use it to collect a photo of a product in use, a short video of a routine, or a document such as a receipt.

• Accepted file types has a toggle for each kind of file. All three are on by default, and at least one must stay on.
• The upload box tells participants which file formats they can pick. After the upload finishes, they see a preview and can Replace the file before pressing Continue.
• Media upload questions have no Ask follow-up questions toggle, because the moderator can't see the uploaded file.

Uploads show in two places. In Results → Analysis, the question's Uploaded files card lists each participant's file: expand a row to preview an image or video, or open or download any file. In the interview's transcript, a View media link sits under the question and opens the file in a new tab.

Matrix

A matrix asks the same question about several items at once. Rows are the items (for example, store names) and columns are the answer options every row shares (for example, "This week", "This month", "Longer ago"). Participants pick exactly one column in each row.

• Add 2 to 10 rows and 2 to 7 columns.
• Randomize rows shuffles the row order for each participant. Columns always keep the order you set.
• Participants must answer every row before they can continue.
• On a wide screen the matrix shows as a table with a radio button under each column. On a phone or beside a stimulus, each row becomes its own box with the options listed underneath.
• In that table, the row text gets twice the width of each answer column, and the table widens on its own when a row is long, so sentence-length rows stay a few lines tall.
• Participants see the question labeled as multiple choice, not as a matrix.
• A matrix answer can't decide which conditional section a participant sees, and can't be used as an analysis filter.

In analysis, the Responses card shows a grid of rows against columns, shaded by how many participants picked each option for each row. Switch between percentages and counts, sort by a column, or filter the columns shown. Each participant's own answers appear as a table of the rows they placed under each option. Renaming a row or column keeps its earlier answers. Deleting one removes its answers from the grid.

NPS

An NPS (Net Promoter Score) question asks participants how likely they are to recommend something on the standard 0 to 10 scale, anchored "Not at all likely" and "Extremely likely". The scale and its labels are fixed, and participants reading in another language see the labels in theirs. For a numeric scale with your own range, use Slider.

• Style sets how the scale looks. Buttons (the default) shows eleven buttons from 0 to 10 that participants tap once. Slider shows a handle they drag to a whole number. Both record the same score, so changing the style never affects answers you already have.
• Participants see the question labeled as multiple choice.
• A later conditional section can depend on the answer, either by group (Promoter, Passive, Detractor), by individual scores, or by comparing the score against a threshold, such as 9 or higher.

In analysis, answers split into three groups: Detractors (0 to 6), Passives (7 and 8), and Promoters (9 and 10). The Responses card shows the NPS score on a gauge from -100 to +100. The score is the percentage of promoters minus the percentage of detractors. Next to the gauge, a 0 to 10 bar chart is colored by group, with each group's share below. In each transcript, Strella tints the score with its group's color.

Best practices

• Use a slider for ratings, scales, and magnitudes (satisfaction, 1 to 7 agreement), and NPS for likelihood to recommend. Reach for multiple choice only when the options are categorical and unordered. See Best practices.
• Don't stack several structured questions in a row: mix open questions in so the interview doesn't feel like a form.
• Match the type to the data: a slider for "how much," multiple choice for "which one," ranking for "in what order."

Tips

• Survey moments interrupt the flow. Use them where structured data matters (NPS, top-3 ranking, satisfaction), not for everything.
• Rating scales with follow-ups are the best of both worlds: a clean number plus a real reason.
• Open text in a survey step is for short typed answers (an email, a one-line description). For longer responses, ask it as an Open question instead: the moderator will get a verbal answer with follow-ups.

Related