Guides4 min read
How to analyze thousands of form responses in minutes
A practical guide to asking Medusa questions in plain English and getting answers you can check against the real responses.
Formzed team
formzed.com
A form with a few dozen responses is easy to read. A form with a few thousand is not, and the useful parts are usually buried in the open answers. This guide shows how to analyze a large set of responses with Medusa, Formzed's AI chat, by asking questions the way you would ask a colleague.
Why reading responses by hand stops working
Most teams start with a spreadsheet. You export the responses, add filters, build a pivot table and scroll through the open answers. That works for a while. Then a form passes a few thousand responses, and every new question means another round of sorting and counting.
Open answers are the hardest part. For example, a customer feedback survey with 2,000 responses might have 1,500 written comments. Nobody reads all of them carefully, so the summary ends up based on whichever comments someone happened to read first.
Medusa takes a different approach. It reads your responses for you, works out the numbers, and shows how it got each answer.
Get set up in three steps
You don't need formulas, filters or a data model. Medusa works from the forms you already have.
- 1Open Medusa and pick your forms. Choose one form, or several if you want to compare them in the same chat. Medusa loads their responses for you.
- 2Ask a question in plain words. Type it the way you would say it out loud, for example "What are customers complaining about the most?"
- 3Read the answer and follow up. Medusa replies with numbers, rankings and charts. Ask a follow-up question to narrow it down, and refer to earlier results with words like "those" or "them".
Medusa reads up to 10,000 responses per question, newest first, so nothing important gets skipped.
Questions worth asking
Not sure where to start? When you open a form, Medusa suggests a few starter questions built from the fields it collects, such as the sentiment in an open question or an average rating by country. They make a good first pass on a new form.
Medusa handles counts, averages, trends, sentiment, themes and rankings. Each answer comes back in the shape that fits it best, such as a bar chart for a ranking or a column chart for a trend. Here are example questions you can adapt to your own forms:
- Counts: "How many people chose the annual plan?"
- Averages by group: "Which countries are most satisfied with our service this month?"
- Trends: "How have registrations trended over the last 6 weeks?"
- Sentiment: "What's the overall sentiment in the open feedback?"
- Themes: "What are the top themes in answers to 'What could we do better?'"
- Rankings: "Rank respondents by their satisfaction score."
- Specific details: "Who are the most experienced candidates that studied at MIT?"
- Uploaded files: "Summarize the uploaded documents."
- Across forms: "Which form gets the most responses?"
Check every answer against the real responses
An AI answer is only useful if you can trust it. Medusa is built to read your data carefully, not to guess.
While it works, you can watch each step live: reading the responses, filtering by a field, grouping by week, and so on. When it finishes, it tells you how long it thought. Answers can point back to the actual responses by their number, such as Response 12, so you can open the original and check it yourself.
Counts, averages and rankings are calculated over every matching response rather than estimated from a few examples. When you narrow a question by country, date or answer, every number recalculates to match. If an answer is based on only part of the data, such as the newest answers, Medusa says so in plain words. When sentiment or themes come back, each theme includes a real quote from a respondent.
A good habit is to ask one thing at a time. A focused question, such as the average rating for one product, is easier to check than a question that asks for five things at once. If an answer surprises you, ask Medusa which responses it used, then open a few of them.
Find the best responses
Analysis isn't only about totals. Often you want to find specific people: the best applicants, the happiest customers, or the users who reported a particular bug.
Medusa can search every answer for a detail and then rank the people who match. For example, you can ask for the five respondents with the highest score who also live in Germany, and Medusa applies both conditions before it ranks. It understands written values like "6-8 years" when it ranks by a number.
If your form collects files, Medusa can read uploaded documents such as PDFs and Word files. You can search inside them, ask for a summary, or rank applicants by their CVs. When you need outside context, such as a public fact about a company, you can turn on web search for a question. Web search answers show the pages they drew on. Each search uses 30 AI credits, plus the credits for reading the results.
The fastest way to see what Medusa can do is to try it on a form you already have. Pick the form, ask the question you would normally answer with a spreadsheet, and compare the result. Every plan includes AI credits for Medusa, including the free one.