What the Survey Response Analyzer does
This analyser reads survey responses from a CSV export and produces the first pass of analysis most surveys need: the distribution of answers to every question, rating-scale summaries, multi-select counts, a cross-tab of any two questions with counts and row or column percentages, and the most frequent terms in open-text answers. Every figure shows its base - how many people answered and how many skipped - so a percentage never hides a small or shrinking sample.
It works with exports from most form and survey tools, because it only needs one row per response and one column per question. The file is read by your browser; responses are never uploaded or stored.
How to use it
- Export responses as CSV from your survey tool, with a header row of question texts. Drop the file on the page, choose it, or paste the CSV into the box, then press Analyse responses.
- Check the question types. Each column is guessed as single choice, multi-select, rating or Likert, number, open text or ignored (IDs and timestamps). Change any wrong guess and the results update.
- Read each question's distribution. Rating questions also get a mean, median and top-2-box and bottom-2-box share; numbers get a mean, median, spread and histogram; open text gets term and phrase counts.
- Pick two single-choice or rating questions for the cross-tab, and switch between counts, row percentages and column percentages.
- Download the results or cross-tab as CSV, or everything as JSON, to take into a report or spreadsheet.
Reading the results
Percentages are of people who answered that question, not of all respondents; the base is printed on every card. For multi-select questions they are of people who answered and add up to more than 100%, because one person can tick several options.
Top-2-box is the share choosing either of the two most positive options - Agree or Strongly agree, or 4 or 5 on a 1-5 scale. It is often easier to explain than a mean score, and it does not pretend the gap between Neutral and Agree equals the gap between Agree and Strongly agree.
Row percentages answer "within each plan, how did people rate us?"; column percentages answer "of people who strongly agreed, which plans were they on?". Pick the one that matches your question, and look at the counts before trusting any difference built on a handful of people.
Worked example: a small product feedback survey
Suppose 10 customers answered, one skipping the plan question. Four were on Pro, three on Free and two on Team, so Pro is 4 / 9 = 44.4% of those who answered - not 40%, which would quietly count the skip as a no.
On "I would recommend it" nine answered: one Strongly disagree, one Disagree, one Neutral, four Agree and two Strongly agree. Scoring 1 to 5 gives a mean of (1 + 2 + 3 + 16 + 10) / 9 = 3.56, and the top-2-box is 6 / 9 = 66.7%.
Cross-tabbing plan by that question uses the 8 people who answered both. Two of the four Pro customers strongly agree - 50% of the row - while none of the three Free customers do. With bases this small, that is a lead to follow up, not a finding.
How question types are detected
Whole numbers that all sit between 0 and 10 are treated as a rating scale (1-5, 1-7 or 0-10, from the largest value seen); other numbers are treated as measurements. Text matching a standard label set - Strongly disagree to Strongly agree, Very dissatisfied to Very satisfied, Never to Always and a few others - is treated as a Likert scale in its natural order, ignoring capitals.
A column is multi-select when at least 15% of answers contain a semicolon or vertical bar and the pieces repeat across responses. A column where nearly every answer is different is open text; a column of unique codes or dates is ignored. These are A2Z heuristics, not a standard, which is why every guess can be changed.
Open-text term counts
Words are folded to lower case, common English stop words are removed and numbers are dropped, then each remaining word is counted, along with how many responses mention it and which two-word phrases repeat. Letters in any script are kept, so Arabic, accented or Cyrillic answers are counted too, although the stop-word list is English only. Counts show what people talk about, not whether they were pleased: "export" could be praise or a complaint.
Limitations: what the result does not prove
- It describes the people who answered. It cannot correct for who did not answer, so results from a self-selected sample may not represent all your customers.
- No weighting and no significance tests are applied. A difference between two groups in the cross-tab may be chance, especially with fewer than about 30 people per group.
- Type detection is a guess. Numeric codes for categories (1 = North, 2 = South) look like a rating scale; set them to single choice.
- Term counts ignore meaning, negation and spelling variants; they point you to answers worth reading, not to conclusions.
Privacy: where your data goes
Everything you paste, type or drop is processed in this browser tab. It is not uploaded, logged, stored or sent to analytics. Session recording and tag-manager scripts are switched off on this page.
Standards and sources
Frequently asked questions
How do I analyse survey results from a CSV file?
Export one row per response with question texts as the header, load the file here and check the detected question types. You get distributions with their bases, rating summaries, multi-select counts, a cross-tab of any two questions and term frequencies for open answers, all downloadable as CSV.
How should multi-select answers be formatted?
Put all of one person's choices in one cell, separated by a semicolon or a vertical bar - for example Search;Export;Reports. That is how most survey tools export checkbox questions. If yours uses one column per option instead, each column can be analysed as a single choice.
What is a cross-tab and when should I use row or column percentages?
A cross-tab counts how answers to one question split across the answers to another. Use row percentages to compare groups defined by the row question, and column percentages to describe who gave each answer in the column question. Always check the counts alongside.
What does top-2-box mean in survey analysis?
It is the share of respondents who chose one of the two most favourable options on a rating scale, such as Agree and Strongly agree. It is widely used because it is easy to explain and less sensitive to how people use the middle of the scale than an average score.
Are blank answers counted as zero?
No. Blanks are counted separately as missing and left out of that question's percentages, averages and cross-tab, and the number excluded is shown. Counting a skip as zero or as the lowest rating would pull averages down and misstate the results.
Is my survey data uploaded anywhere?
No. The CSV is read and analysed by your browser and never sent to a server or saved. Closing the tab discards it. Downloads are generated in the browser, and exported cells that start with =, +, - or @ are prefixed so a spreadsheet will not run them as formulas.
Last reviewed by the A2Z.Tools team against the sources listed above.
