Reading responses

Open a survey from the list to see its results. Pick a date range and narrow by device, browser, country or segment; every tab respects the same filters.

Overview

• Impressions, starts, completions and completion rate — how many people saw the survey versus how many finished it.
• Median time to complete — a long one usually explains a low completion rate.
• NPS, CSAT and effort score, shown only when the survey actually asks a question of that kind.
• A trend chart of impressions against completions, and a country breakdown.

Questions

One card per question. Rating and choice questions aggregate into distributions with averages; NPS questions break down into promoters, passives and detractors; open-text questions list recent answers.

Above the cards, What people are telling you summarises every written answer: recurring themes with their sentiment, the requests that came up most, and an overall sentiment split. Each theme also appears as a chip on the question it came from. The summary is saved rather than recalculated on every visit, so it shows when it was last analysed — press Re-analyse after a batch of new responses.

Drop-off

The funnel goes shown → started → answered each question → completed, so you can see the exact question people quit on. Partial answers count: someone who answers three of five questions and leaves still contributes everything they did answer.

Abandoned sessions link straight to the session replay at the moment the survey was on screen, which is usually faster than guessing why the question lost people.

If the survey uses branching, remember that a question skipped by a rule shows fewer answers without anyone having abandoned it — a dip in the funnel is not always a problem.

Responses

Every response with its answers, one row each, exportable to CSV. Open a row for the full detail, or delete a single respondent's answers if they ask you to.

Tips

• Open-text answers are good fuel for A/B test ideas and copy changes.
• If you are running variants, the significance table tells you whether a difference in completion rate is real or noise.
• Low completion with high impressions is usually a length problem, not an interest problem. Cut questions before you rewrite them.