Surveys & NPS
Ask your users a question inside your product and get the answer back where you already read your onboarding numbers. NPS, rating scales, choose-from-a-list and written answers - on every plan, including free.
How a survey is put together
A survey is a flow type, so everything you already know applies: targeting decides who sees it, frequency decides when, and appearance decides how it looks. What is different is the content - a survey's steps are questions.
- Create a flow and pick Survey as the type - or use Generate with AI and describe what you want to learn, and it drafts the questions with the right kinds (a score, a rating, a choice list, an open answer).
- On the Questions tab, start from a preset (NPS, CSAT, CES) or add your own.
- On Appearance, choose where the card appears; on Frequency, how often the same person may be asked.
- Publish. Answers appear on the flow's Analytics as they arrive.
Question types
| Type | What it collects | How it reports |
|---|---|---|
| NPS | A score from 0 to 10 | Your NPS number, the promoter / passive / detractor split, and the spread across all eleven scores |
| Rating scale | A score from 1 to N | An average and a distribution |
| Choose from a list | One option, or several if you allow it - optionally with an “Other” box | A count and share per option |
| Written answer | Whatever they write | The answers themselves, newest first |
Rating scales
A rating can be drawn four ways - numbers, stars, faces, or labelled ends. This is presentation only: the answer is stored as a number either way, so you can switch from stars to faces mid-survey and your existing answers stay comparable.
NPS is always 0-10. That is what makes the score comparable with everyone else's, so the scale and length are not adjustable.
Choosing more than one
A list question takes one answer by default and submits as soon as it is tapped. Turn on Let them choose more than one and the card stays open while options are toggled, with a Done button to submit.
Each selected option is recorded on its own, so the report counts every option independently. That means the per-option counts on a multi-select question can add up to more than the number of people who answered - which is correct, and worth knowing before you read the percentages.
CSAT, CES and other methodologies
These are presets, not separate types - a CSAT is a rating with a particular scale and wording. Picking a preset fills in the question for you; you can change anything afterwards.
Where the card appears
| Position | Behaviour | When to use it |
|---|---|---|
| Any corner | Slides in at that corner, page stays usable | The default. Easy to ignore, which is the point for a recurring NPS. |
| Centre | A modal, with a backdrop | Gets the most answers and interrupts the most. Right for a deliberate, occasional sweep. |
| Beside an element | Anchored to something on the page | When the question is about a specific thing: “how was checkout?” next to checkout. Overrides the position. |
Position is set on the Appearance tab beside the live preview, so you can see the effect as you change it. The same tab carries the entry animation, and the card's colours, font, text size, corner radius and shadow.
For the anchored option, give a CSS selector - or capture one with the builder extension, which produces a selector that survives a redesign. If the element is not on the page when the survey runs, it appears in the corner instead rather than not at all.
Asking again
How often the same person may be asked is set on the Frequency tab, the same as any other flow. For a recurring NPS, Show again after an interval of about 90 days is the usual choice - long enough that people are not nagged, short enough to see a trend. Leave it on Once and nobody is ever asked twice.
Going back
From the second question onward the card offers Back, so someone who mis-taps can correct it. Their corrected answer replaces the first rather than adding to it, so one person is only ever counted once per question.
“Ask me later”
Turning this on (in Appearance) adds a snooze to the first question. A snooze is treated as a genuine “later”: the person is asked again in about a week, and then the normal frequency rules take over.
Each answer is recorded the moment it is given, not at the end. Someone who gives you a score and then closes the card has still given you their score. That is why a survey's later questions usually show fewer answers than its first - both numbers are true, and the results show each question's own count.
Reading the results
A survey's Analytics leads with the answers. Each question is reported in the way that suits it: a score and split for NPS, an average and spread for ratings, per-option counts for options, and the answers themselves when people wrote something.
The 0-10 spread is worth a look even when the headline score is good. A +40 built from a wall of 9s is a different product from a +40 built from a fight between 10s and 3s, and the score alone cannot tell you which you have.
The headline numbers describe a survey on its own terms: how many people were shown it, how many answered, and how many answered every question. How far people got is the survey's drop-off - if most people answer the first question and few reach the last, the survey is too long or a question in the middle is doing the damage.
Export CSV on the Answers panel downloads the raw responses for whatever date range and version the page is showing: one row per person, one column per question, with option labels and written answers in full rather than the numbers they are stored as. Export is on the paid plans.
Plans and privacy
Surveys are on every plan, including free, where they carry the same Onbixo badge as free-tier flows. There is no per-answer charge; your MAU cap applies exactly as it does to any other flow.
Answers are kept indefinitely - they are not rolled up or deleted on the schedule that applies to analytics events, because an NPS trend across two years is the point of running NPS. They are deleted when you erase an end user, when you delete the survey, or when you close your account.
Related
- Targeting - decide who gets asked.
- Frequency & triggers - when a flow shows.
- A/B testing - test one wording of a question against another.
Last updated: Sep 2, 2026