Survey Results Analyzer

About the Survey Results Analyzer

Open the CSV or Excel export of a form or survey — one row per response, one column per question, as Google Forms, Microsoft Forms, Typeform and most survey tools export it — and every question is summarised at once. The analyzer works out what each column is: a single-choice question, checkboxes (several answers per person), a Likert item (strongly disagree … strongly agree, very dissatisfied … very satisfied, never … always), a rating scale, a 0–10 recommend question for the Net Promoter Score, a number or open text. Timestamps, IDs, names, e-mail addresses, phone numbers and the links of file-upload questions are left out (you can still give any of them a type).

Each question gets its counts and percentages with a chart: bars for choices, diverging stacked bars for Likert items with the median answer and the share at the top and bottom of the scale, the Net Promoter Score with its margin of error, quartiles and a histogram for numbers, and the words most answers use for text. Choose a question such as department or plan to break every other question down by it, with a chi-square test of whether the groups really differ. Download the summary as Excel or CSV, every chart as PNG or SVG, or print the whole report or save it as a PDF. The file never leaves your browser.

Common uses

• Customer satisfaction and feedback surveys: satisfaction by plan or region, NPS, and what people ask for in their comments.
• Employee engagement surveys with Likert statements, broken down by team or department.
• Course and event feedback forms collected with Google Forms or Microsoft Forms.
• Student research projects and questionnaires that need frequency tables and cross-tabs.
• Turning a form export into a summary workbook or a PDF report to share.

How each kind of question is summarised

• Single choice: the count and share of every answer among the people who answered. Answers given only once in a question that is otherwise a few fixed options (what people type under “Other”) are grouped as Other (written in) and listed.
• Multiple choice (checkboxes): the share of people who chose each option. Several answers in one cell — separated by commas, semicolons or line breaks — are split, and so are exports that give every option a column of its own.
• Ranking: when every answer lists the same options in an order (“Reports;Invoices;Chat;”), the average position of each option (1 = first) and the share of people who put it first.
• Likert items and other ordered scales: the share of every answer in scale order, the median answer, the share in the top two and bottom two answers, and a mean score for reference. Grid questions exported as one column per row, headed “Question [Row]”, are shown together.
• Ratings (whole numbers on a short scale, such as 1–5 stars): the distribution, mean and median.
• Net Promoter Score: from a 0–10 question about recommending.
• Numbers: count, mean, standard deviation, minimum, quartiles (Excel’s QUARTILE.INC method), maximum and a histogram.
• Open text: every answer, and the words that the most answers use (common English words such as the and and are left out).

Likert items and diverging bars

Rensis Likert’s A Technique for the Measurement of Attitudes introduced the five ordered answers from strong approval to strong disapproval that survey tools still offer, now in 4- to 7-step versions. The steps are in order but not known to be equally far apart, so the median and the shares of each answer are the safest summary; the mean score (1 for the lowest answer up to 5 or 7) is shown because many reports use it.

In the diverging bar the negative answers run left of the middle line and the positive ones right of it, with a neutral answer split across the line, so that several statements or groups can be compared at a glance. Words the analyzer recognises include agreement, satisfaction, likelihood, ease, helpfulness, expectations, quality, frequency, importance and familiarity scales, and answers numbered “1 – Poor” … “5 – Excellent”. When nobody chose the middle answer the scale is read without one.

Net Promoter Score

For a 0–10 question about recommending (How likely are you to recommend us to a friend or colleague?), answers of 9 or 10 are promoters, 7 or 8 passives and 0 to 6 detractors; the score is % promoters − % detractors, from −100 to +100, as Fred Reichheld described it in The One Number You Need to Grow. The margin of error at 95% confidence is 1.96 × 100 × √((p₊ + p₋ − (p₊ − p₋)²) ÷ n), with p₊ and p₋ the shares of promoters and detractors and n the number of answers. Net Promoter®, NPS® and Net Promoter Score℠ are trademarks of Bain & Company, Inc., NICE Systems, Inc. and Fred Reichheld; the names only describe the calculation here.

Cross-tabs and the chi-square test

A breakdown shows, for every group, the share of the group that gave each answer (column percentages), so groups of different sizes can be compared. For single-choice and Likert questions the chi-square test of independence checks whether the differences are bigger than chance would produce if the answers did not depend on the group: χ² = Σ (observed − expected)² ÷ expected, with (rows − 1) × (columns − 1) degrees of freedom, as in the NIST/SEMATECH e-Handbook of Statistical Methods. A p-value below 0.05 means a difference that large would be unusual by chance alone; Cramér’s V (0 to 1) says how strong the link is.

Two cautions. The test is rough when many expected counts are small — the tool says so when more than 1 in 5 are below 5. And with many questions, about 1 in 20 tests will come out below 0.05 by chance even when nothing differs, so treat a single small p-value as a lead to check, not proof. Multiple-choice questions get no test, because one person can be counted under several answers.

Limitations

• Files up to 100 MB as CSV or 50 MB as a workbook; the summaries are worked out in your browser’s memory.
• The first row must hold the questions. A second heading row that says “Response” or names each option under its question is combined with the first automatically; other layouts may need tidying first.
• A checkbox option that itself contains a comma followed by a capital letter (“Yes, Definitely”) is split in two when options are separated by commas; choose another type for that question if needed.
• Responses are not weighted, and the tests assume a random sample: people who chose to answer may not speak for everyone.
• The word list knows common English words only; for other languages the most used words may include words such as “and”.

Privacy

Everything happens in your browser. What you enter or open here is not uploaded or stored by MySmartCoPilot.