Analytics
Per-question analysis
What each question type reports.
Open a survey's Analytics, then choose Questions. Select a question to inspect its answers and the analysis available for that question type.
Read the right summary
| Answer type | What to look for |
|---|---|
| Choice questions | Counts and proportions for each option |
| Numeric and scale questions | Distribution, centre and spread |
| Text questions | Aggregated word frequencies, sentiment and other available text summaries; use Responses to read individual answers |
| Ranking, Likert and Matrix | Results for the individual items, statements or rows |
Question analysis defaults to completed responses. Use Filters to change the status selection or date range. A question can have fewer answers than completed responses because it was optional or skipped by a branch.
- Select the question in the left-hand list.
- Check the number answered before comparing results.
- Read the most common answer and its count alongside the chart.
Check the underlying answers
A summary can hide outliers or missing answers. Open the Responses view when you need to understand a particular submission before drawing a conclusion.
See filtering, statistics glossary and exporting.
Example: follow-up ratings after a branch
Suppose 100 people complete a workshop survey, but only attendees see a rating question. If 60 attended, that rating should not be interpreted as a result from all 100 respondents. If the rating is optional, its answer count may be lower still.
Read the question's answer count before its percentages or averages. Then check which route reaches that question and whether answering was required. This helps distinguish a smaller intended group from an unexpected omission.
Interpret the question's structure
For a single-choice question, one respondent contributes one selected option. A checkbox question can contribute several selections, so counts across options need not add up to the number of respondents. For a Matrix or Likert question, inspect the relevant row or statement rather than treating the whole question as one scalar answer.
For numeric answers, read the distribution alongside any summary value. The same average can come from answers clustered near the middle or split between low and high values. For text, read actual answers before treating a frequent word or sentiment summary as a complete explanation.
If the result looks empty or unexpected
Check the response status filter, then the date range. Confirm that the question was shown on the route you are inspecting. Open a known submission in Responses to compare its answer with the summary. Keep those checks separate from changing or deleting the source response.