RawQuotes

“What you see is all there is.” Daniel Kahneman · Thinking, Fast and Slow · 2011

Customer feedback grouped into insights you can check.

Put a short survey on any page, or bring what you already have: a Typeform export, your Google Play reviews, last month's support e-mails. RawQuotes groups it into named insights, and every insight opens into the exact words it was counted from. Under ten complete responses, you get a count instead of a guess.
Create account

14-day trial, MCP included, no card

See where feedback can come from

Feedback from four places, unsorted

Survey Wrote on Monday, got an answer on Thursday

Google Play Good app, but nobody answers the help form

E-mail Three e-mails to fix one typo on my invoice

Typeform Signed up but couldn't tell what happens next

Named insights, with their quotes

Live example · demo data

Open an insight. Read the words it was counted from.

64 pieces of feedback from four places: a survey on the site, a Typeform export, Google Play reviews and support e-mails. RawQuotes read them together and named four insights. Click one.

Demo project: a mobile planner app · Last 30 days, 5 Oct to 3 Nov 2026
GeneratedProblem

Sign-up asks for too much before people see the app

People want to see the app before they give company details, a phone number or confirm an e-mail.

14 answers · 22% Up 7 ppvs previous 30 days

Backed by 14 from 4 sources

Survey 6 Google Play 3 E-mail 2 Typeform 3

Respondents' own words

SurveyWhat almost stopped you from signing up? · 28 Oct
“Asked for company details before showing a single screen”
TypeformOnboarding survey, October · 24 Oct
“Ten fields in before I saw any actual feature”
Google PlayGoogle Play reviews · 26 Oct
“Wasn't sure what I'd get before creating an account. The store screenshots don't show much either”
E-mailSupport e-mails · 14 Oct
“Hi, I gave up halfway through sign-up yesterday. Can you just send me a demo login?”
26 answers did not fit a repeating insight

Insight names and descriptions are written by the model. Everything else here, every quote and every count, comes from the people who wrote in.

Demo project with made-up feedback written for this page. Not a customer.

Try it on your own feedbackYour agent gets the same quotes over MCP

Feedback piles up in four places, and nobody reads it

Survey answers sit in one tool. A form export sits in a spreadsheet, app reviews in Play Console, support e-mails in an inbox. Reading all of it takes an evening nobody has. After that you still have to group it, count what repeats, and set the one-word replies apart from the ones that mean something. So the reason your numbers moved stays unread.
See where people already tell you why

How it works

01

Collect it, or bring it in

Put a short survey on any page with one script tag. Or bring feedback you already have: upload a CSV, Excel or JSON export, paste text, or send it through the API from n8n, Zapier or Make.

02

Read together, by meaning

Everything in a project is read as one set. A complaint in a review and the same complaint in a survey can land in one insight. Every quote shows the source and the question it came from.

03

Ten responses before any insight

Once a project has ten complete responses in a period, matching feedback is grouped and named in one sentence: a problem, a piece of praise, or a reason. One imported review, e-mail or row counts as one response. Below ten, you get a count instead of a guess.

Analyse feedback from any source you already have

Run a survey on your site, import an export, paste a batch of e-mails, or connect the tools you already use through n8n, Zapier or Make. Everything lands in one project and is read together.
Connected through n8n, Zapier or Make and the RawQuotes API.
How each source works

Check every insight against the words behind it.

Every insight opens into the answers it came from

An insight is one sentence naming what people keep saying, with a count next to it. Open it and you get the feedback it was counted from, word for word, with nothing shortened or paraphrased. The list is exactly as long as the number says. Each quote shows its source and the question it answered. The insight's name and description are the only text there nobody wrote to you, and both are labelled as generated.

Panel screenshot: the insight "Delivery took longer than the date shown at checkout", labelled Generated and Problem, backed by 16 items: 7 from a checkout survey, 5 Google reviews and 4 support e-mails. It is opened into three verbatim quotes, one from each source. Under each quote are its source, the survey question it answered ("What should we improve?") or the source's description, and its date, followed by a link to see all 16 answers

Under ten complete responses, you get a count instead of an insight

Grouping a handful of answers produces something that looks convincing and means nothing. So below ten complete responses this app shows no insights at all. It shows how many you have, how many it needs, how many are missing, and a link to the raw replies. One imported review, e-mail or spreadsheet row counts as one response, the same as one finished survey. Filter by one source, and an insight with fewer than three items from it says "below threshold for this source" instead of a number.

Panel screenshot: a refusal below the threshold of ten complete responses, titled "4 more complete responses to name your insights", with a progress bar at 6 of 10

When a recalculation fails, it says so

Insights are recalculated on a schedule. If a run fails, the panel says it failed and shows the last successful result with the date it was calculated — never a stale number pretending to be fresh, never an insight without a date. Your answers and filters keep working through it.

Two of those three are refusals, and that is the point. A tool that never says "I don't know" gives you no way to tell a real pattern from a filled screen. It also means there will be weeks when this app shows you nothing — we would rather do that than invent an insight.

See whether a problem is shrinking or growing

Pick a week or a month and compare it with any earlier one. Every insight shows its share of answers in both and how many percentage points it moved. A problem that grows is marked red and one that shrinks green; for praise it is the other way round, and insights that are neither stay grey. Each share is counted against that period's own responses, so a busier month doesn't read as a worse one. If fewer than three people mentioned an insight in the earlier period, it says "Not compared" instead of showing a number.

Panel screenshot comparing Period 2 with Period 1 (120 vs 100 responses): a "What changed" summary naming the biggest rise, a delivery problem up 9 pp, and the biggest fall, a sizing problem down 8 pp; below it two insight cards with their share in both periods — the delivery problem up 9 pp in red and gift-ready packaging praise up 6 pp in green

New: MCP server

Connect your agent

  1. 1

    Add the server to your agent.

  2. 2

    Approve what it may do.

  3. 3

    Ask. It reads insights with the quotes and responseId behind them, or hears “too few” under ten.

Basic and Pro plans, also in the 14-day trial

How it works with your agent
Simulated chat: asked "What do people keep telling us about Planner app? Quote them.", an AI assistant calls list_projects and get_insights, then names the top insight, "Task changes are not synced reliably between devices" (6 responses), with two word-for-word quotes, each followed by its responseId.
Example data. Simulated chat view; the quotes and response IDs come from a real get_insights call.

Before anything reaches the AI

Sensitive details get masked before any feedback reaches our AI provider

People type an e-mail, an IBAN or a phone number into a survey without thinking, and support e-mails carry them in every signature. On our own server, before any text reaches our AI provider, we replace those details with a placeholder. The original stays in your account, and every quote you read comes from it. Masking is best-effort and misses some things: names, for example, are not masked. It still helps you meet GDPR's data-minimisation principle for the text we send to our AI provider. An agent you connect yourself gets the original text, and only if you allow it.

See what we mask, and what we don't

What leaves your account, and what the AI model sees, is in the privacy policy

Feedback

Shipping was fast, but nobody replied when I emailed kontakt.klient@example.com about a discount code for order order-00048.

Pricing

Free, Basic and Pro. Each plan includes everything in the one below it and raises the limits: projects, surveys, monthly responses, seats and how often insights are recalculated. Yearly billing costs two months less.

Every plan runs on the same insights engine, wherever the feedback came from: quotes under every insight, and no insight below ten complete responses.

2 months free
Free
One survey, quote-backed insights, free forever.
Free
  • 100 responses per month
  • 1 active survey
  • 50 analyzed answers per month
  • 1 user
  • 1 project
  • 5 recalculated insights per project
  • 1 custom tracked insight
  • 90-day data retention
See pricing
No card required.
Basic
14-day trial
Everything in Free, five times the surveys, with a 14-day trial.
$29
billed monthly + VAT/month
  • Everything in Free, plus:
  • 5,000 responses per month
  • 5 active surveys
  • 500 analyzed answers per month
  • 3 users
  • 3 projects
  • 10 recalculated insights per project
  • 5 custom tracked insights
  • Manual insight recompute, each run using analyzed answers from your monthly pool
  • 360-day data retention
  • Custom metadata on answers (coming soon)
  • API, to send feedback from n8n, Zapier, Make or your own code
  • MCP access
  • 14-day free trial with 100 analyzed answers
See pricing
Cancel any time.
Pro
Everything in Basic, fifty surveys, refreshed daily.
$69
billed monthly + VAT/month
  • Everything in Basic, plus:
  • 100,000 responses per month
  • 50 active surveys
  • 2,000 analyzed answers per month, up to 50,000
  • 20 users
  • 20 projects
  • 10 recalculated insights per project
  • 50 custom tracked insights
  • Manual insight recompute, each run using analyzed answers from your monthly pool
  • 3-year data retention
  • Custom metadata on answers (coming soon)
  • API, to send feedback from n8n, Zapier, Make or your own code
  • MCP access
See pricing
Cancel any time.

For businesses only. Cancel any time. Prices exclude VAT and other sales taxes. 1 AO = up to 1,000 characters of one answer

Common questions

How is this different from other feedback tools that group answers with AI?

You can check every insight. It opens into the exact feedback it was counted from, nothing is named under ten complete responses, and a failed recalculation is shown as failed, never as a fresh result.

Can I analyze feedback I already have?

Yes. Upload a CSV, Excel or JSON export, paste text, or send it through the API from n8n, Zapier or Make (Basic and Pro). It's read together with your surveys, under the same rules.

Why am I not seeing any insights?

There are fewer than ten complete responses in the period, counted across all your sources. You get a count instead of a guess: how many you have, how many are needed and how many are missing.

Do you train AI on my feedback?

No. Survey answers and imported feedback go to our AI provider only to be grouped, and neither we nor the provider train any model on them.

How much does it cost?

Free, Basic and Pro, from $0. Every plan is a superset of the one below it — see the full pricing page.

See all questions

See what people keep telling you.

Create account