25 Questions for a Travel Agency Customer Satisfaction Survey Comparing with past results
Compares a survey with its own past, because one figure means little and a change says a lot
Free agent skill for Claude Code, Codex, Cursor and Gemini CLI. It works through the SurveyNinja MCP server, with your permissions.
- Category
- Understand results
- Tags
- comparison, report, edit history
- Works with
- Claude Code, Codex, Cursor, Gemini CLI
- MCP tools
- 7, all read only
- Version
- 1.1, October 20, 2026
- Made by
- SurveyNinja
- Price
- Free
- Skill file
- SKILL.md
What it does
- You see whether things got better. A rating of 4.2 says little on its own. It means something next to the previous measurement.
- Comparability comes first. If the survey was edited between waves or the periods differ in length, the skill says so instead of showing a trend.
- Guesses are labeled. The skill names a cause only when the data shows it, and marks any guess as a guess.
Example
You type
Compare the autumn wave of the survey with the spring one
What the assistant brings back
- 4.0
- Spring286 responses
- 4.3
- Autumn318 responses
- +0.3
- Change
- unchanged
- Surveyno edits between waves
The average moved three tenths, but the interesting part is elsewhere: the very dissatisfied group halved and the very satisfied group grew noticeably. The middle barely moved.
The numbers above are an example. The skill compares your own periods and surveys.
How it works
- Checks comparability firstIf the wording or answer options changed between waves, the numbers are not comparable. The skill says so plainly instead of showing a "trend".
- Looks at the distribution as well as the averageThe average can stay flat while half the satisfied moved to dissatisfied and the other half moved the other way.
- Separates the two kinds of comparisonOne survey across periods is reliable. Different surveys can only be compared on shared metrics such as completion rate and time to complete.
- Says whether the change mattersOn small samples a tenth of a point means nothing, and the skill says so instead of calling it growth.
Try also
- Did things get better or worse since the last measurement?
- Compare completion rates across two surveys in my account
What it will not do
- Do not compare periods of different length
- Do not compare waves with an edited questionnaire
- Do not compare seasonal periods head-on
- Do not suggest a plan upgrade or lead to payment
The skill works with your permissions. Tools that change data are marked, so you check what the assistant is about to do before you confirm. To cut access, remove the connector in the assistant or delete the API key.
In short
What can I compare?
One survey across periods, in waves, which is the most reliable. Or different surveys in your account, but only on shared metrics: completion rate and time to complete.
What does the skill check before comparing?
The publication history: if the wording or answer options changed between waves, the numbers are not comparable and the skill says so. It also checks that the periods are the same length.
Why is the average not enough?
The average can stay the same while half the satisfied respondents became dissatisfied and the other half went the other way. The distribution across options shows where a shift came from.
How will I know whether things improved?
From the direction and size of the change, such as "the average rose from 4.0 to 4.3", with how many responses each figure rests on, and whether the difference matters on samples of that size.
Will it explain why the numbers changed?
Only if the reason is visible in the data. Otherwise it offers a guess and labels it as one.
Does the skill change anything in my surveys?
No. It only reads the survey list, the summary, the report, the responses and the publication history.
The full skill
The SKILL.md file the assistant reads, exactly as it is in the archive.
Show the full text
Comparing with your past self
A single number means almost nothing: is 4.2 good or bad? Meaning appears only against a previous measurement.
Reply to the person in the language they write in.
Two kinds of comparison
One survey across periods, in waves. The most reliable: same questionnaire, same audience, only time differs.
Different surveys in the account are comparable only on general metrics such as completion rate and time to complete. Substantive answers cannot be compared when the questions differ.
How to collect
For waves: get_quiz_report with date_from and date_to, once per period. The
dates work only when the plan has results filters (result_filters in
get_workspace_tariff, with the workspace_id from get_workspace_list);
otherwise the report silently covers all time, and two periods give identical
numbers. Without the filters, read the completed responses with
get_quiz_answers (is_complete true, paged with limit and offset) and split
them by their end date, or say that waves cannot be compared here. The report
holds only completed responses either way; completion per period comes from the
daily counts in get_quiz_summary (number_answer_by_day; the unfinished ones
appear only when the plan keeps them).
For different surveys: get_quiz_list, then get_quiz_summary for each. The
summary gives visits, completed and abandoned responses, completion rate and
average time.
Check get_quiz_versions before comparing waves. Every publication leaves a
snapshot with its date. If the survey was republished between the waves and the
wording or options changed, the numbers are not comparable, and that must be said
plainly rather than shown as a trend. If the snapshots do not make clear what
changed, ask the person.
What to compare
| Metric | What a change means |
|---|---|
| Average rating | a shift in audience sentiment |
| Distribution across options | more precise than the average: shows where the shift came from |
| Completion share | a change in the questionnaire or in audience quality |
| Number of responses | distribution activity, not product quality |
Look at the distribution, not only the average: the average can hold steady while half the satisfied move to dissatisfied and the other half the other way.
How to answer
- State direction and size. "The average rose from 4.0 to 4.3", and how many responses each is based on.
- Say whether it is meaningful. On small samples, tenths mean nothing.
- Do not explain a cause that is not in the data. A guess can be offered, but labeled as a guess.
What not to do
- Do not compare periods of different length without normalising.
- Do not compare waves with an edited questionnaire in between without saying so.
- Do not compare seasonal periods head-on. December and July behave differently.
- Do not suggest a plan upgrade or lead to payment. If a limit is hit or a feature is not on the plan, state the fact and stop.
More skills
All skills- A/B variant analysis Tells you whether the difference between variants is real or just noise
- Survey results summary Answers "so what did the survey say" with figures and a conclusion instead of a raw export
- Completion funnel Shows which question makes people close the tab, and why
- Survey timeline Answers “why did responses stop” by lining up collection against the edit history
- How many responses you need Tells you whether what you have is enough and how many more responses you need before conclusions hold
- Analyze open-ended answers Turns a hundred lines of free text into a handful of themes with numbers and quotes
Connect SurveyNinja to your assistant
The MCP server works on every plan, Free included. Sign up, allow access and ask your assistant to build the first survey.