Feedback you don't read still costs you.Feedback you act on pays.
Sentriment reads every review, ticket and survey, prices each theme in the MRR behind it, warns you which accounts are drifting toward churn, and shows you which users love you, and why, in their own words.
No credit card · Set up in 5 minutes · 100% EU-hosted, GDPR-first
- apiusr_8231 · 14:02
- received“checkout froze when I applied my coupon”
Every channel, one pipeline.
POST from any backend, drop in the widget, or connect the tools your feedback already lives in. Every source is read the same way, and what comes out lands where your team already works. Read the docs →
REST API
liveOne POST from any backend. The item is redacted on arrival and joins its user's record.
$ curl -X POST \
https://app.sentriment.com/api/v1/feedback \
-H "Authorization: Bearer sk_live_…" \
-H "Idempotency-Key: order-4231-note" \
-d '{
"text":
"checkout froze when I applied my coupon",
"userId": "usr_8231"
}'Any language, one call
Send one item or thousands. Idempotency keys make retries safe.
Users come with it
Pass a user id and the item feeds that user's health score.
Errors you can read
Every failure is RFC 7807 problem+json, with a reason.
coming next
Slack & email alerts
Spikes, regressions, and at-risk users, the moment they happen
Mixpanel & Amplitude sync
Health scores pushed onto user profiles. Cohort and campaign on them
GitHub & Linear issues
File from a theme; feedback linked to the fix
Daily digest
What changed overnight, every morning on your Insights page
Dashboard, API & webhooks
Query clusters programmatically, push events out
REST API
liveOne POST from any backend. The item is redacted on arrival and joins its user's record.
$ curl -X POST \
https://app.sentriment.com/api/v1/feedback \
-H "Authorization: Bearer sk_live_…" \
-H "Idempotency-Key: order-4231-note" \
-d '{
"text":
"checkout froze when I applied my coupon",
"userId": "usr_8231"
}'Any language, one call
Send one item or thousands. Idempotency keys make retries safe.
Users come with it
Pass a user id and the item feeds that user's health score.
Errors you can read
Every failure is RFC 7807 problem+json, with a reason.
Slack & email alerts
Spikes, regressions, and at-risk users, the moment they happen
Mixpanel & Amplitude sync
Health scores pushed onto user profiles. Cohort and campaign on them
GitHub & Linear issues
File from a theme; feedback linked to the fix
Daily digest
What changed overnight, every morning on your Insights page
Dashboard, API & webhooks
Query clusters programmatically, push events out
coming next
One pipeline, whatever the source
A review, a ticket, a note from the widget and a row from a CSV are redacted, deduped and scored the same way.
One person, one profile
Email joins a customer's Intercom chats, Zendesk tickets and in-app notes into one record with one health score.
History without the noise
Backfills and CSV imports join themes and score health, but they never fire alerts or register as spikes.
- Exporting tickets to a spreadsheet
- Reading reviews by hand
- apiusr_8231 · 14:02
- received“checkout froze…”
- redacted3 identifiers → ⟨name:1⟩ ⟨email:1⟩ ⟨phone:1⟩
Personal data never reaches storage.
Names, emails, phone and card numbers become typed tokens on the way in. The raw text never touches the database, and the database is in Frankfurt. How redaction works →
Redacted before storage
Typed placeholders replace PII before a single byte hits the database, so there is no raw PII to breach, subpoena or leak into a model.
100% EU data residency
Database in Frankfurt, compute and backups in the EU; nothing crosses the Atlantic, at rest or in processing.
Your data trains nothing
Feedback is processed to serve your dashboard, never used to train third-party models; every AI output records the model and prompt that produced it.
- Sending raw customer text to a US vendor
- A DPA that hopes for the best
# in-app widget
"refund me at sara.k@acme.io, card 4242 4242 4242 4242"
received…
# POST /api/v1/feedback
"Marta Ellis, call me on +34 612 04 88 91"
received…
- PII redacted
- EU-hosted
- GDPR-first
Right to be forgotten, built in
One call erases a user's feedback and identity everywhere. Clusters recount, exports comply. Deletion is a feature, not a ticket.
DELETE/api/v1/users/usr_8231
204 No Content · cascade completed
4 identifiers removed · processed and stored in the EU · never used to train third-party models · erased in one call
- apiusr_8231
- redacted3 identifiers
- clusteredCheckout freezes on coupon · 42 reports · €4,180
See the revenue behind a complaint.
Every item is read, deduped and grouped by meaning. Each theme carries the MRR of the accounts that reported it, so the roadmap argument is about money, not volume. See a theme →
Themes
last 30 days · 1,284 items
ranked by MRR at stake · 1,284 items · last 30 days
Checkout freezes on coupon
▲ 340%€4,180
MRR behind this theme · 42 reports
“Coupon field just spins forever on my iPhone.”
“Cart freezes the second I apply SAVE20.”
42 items cited · first seen 6h ago · checkout v4.2.1
Priced in MRR
The revenue behind a theme is correlated from the identity traits you already send, no extra tracking.
Spikes surface within hours
A theme rising 340% in a day is flagged and summarised before the next review.
Quotes stay attached
Every count links back to the verbatim items behind it, so nothing is paraphrased away.
- Manual tagging in Zendesk
- Vote counts in Canny
- apiusr_8231
- clusteredCheckout freezes on coupon
- scoredusr_8231 · 23 · at risk
Catch churn while it is still preventable.
Each user gets an explainable health score from the tone and weight of what they send. A sustained decline fires an alert before the renewal call, not after it. How the score works →
Users at risk
score < 40 · 7-day change
User at risk: usr_8231
Slack · #cs-alertsscores recomputed nightly and on new feedback · synced to Mixpanel and Amplitude
7 items weighed · sentiment −0.42 · emotion penalty −12
Ranked by MRR
Risk is sorted by what the account pays, so the biggest exposure comes first.
Alerts where you work
Slack or email the moment someone drops into at-risk, with who, why and what they pay.
Synced back
Scores land in Mixpanel and Amplitude as user properties, so cohorts and campaigns can use them.
- Finding out at the renewal
- A churn dashboard that reports the past
- clusteredCheckout freezes on coupon
- scoredusr_8231 · 23 · at risk
- actedalert · #cs-alerts · filed SEN-142
Ask a question. Get a cited answer.
Ask in plain language across every source at once. The answer carries a verdict, a number and the quotes it rests on, and a tracked question re-answers itself on a schedule. Ask and tracked questions →
Cited and auditable
Every answer records the model and prompt version that produced it, so every claim in your dashboard can be checked.
Alerts that matter
Slack or email when a theme surges, a fixed issue returns, or a user drops into at-risk.
Out to where you work
File a theme to GitHub or Linear, get the daily digest, push events over webhooks or pull them from the API.
- The weekly 'what are users saying' deck
- Searching the support inbox
“We can't roll this out without SAML.”
“Okta or it's a non-starter for our security team.”
claude-opus-4-8 · prompt 2026-07-19.3 · re-answered 2h ago
also in the box
- semantic search
- saved reports
- trait segments
- triage workspace
- “You asked, we shipped” notes
- CSV import & backfill
- outbound webhooks
- public REST API
every answer stamped with its model and prompt · tracked questions re-answer on a schedule
- widgetusr_2210 · 09:41
- received“love the new editor!!”
- praisedEditor praise · usr_2210 · 91 healthy
Find the customers you should clone.
The same score that flags churn ranks your happiest accounts, and praise is clustered like complaints: counted, cited and priced. The words your best users use are the words your next users will recognise. Segments and traits →
Who loves you, by MRR
Healthy accounts ranked by what they pay and what they praise, so the ideal profile is a list, not a guess.
What they love, in their words
Praise themes carry the verbatim quotes behind them, ready for a landing page, an ad or a sales deck.
Which segment loves which feature
Filter by plan or trait to see where the praise concentrates, and where it does not.
- Guessing your ICP
- Testimonials you have to ask for
Users who love you
score ≥ 70 · by MRR“love the new editor!!”
“onboarding was actually fun?”
praise clustered like complaints · quotes verbatim · scores recomputed nightly
Measured
From raw quote to ranked theme in 8.7 seconds.
Three numbers the site publishes because they were measured, not promised. Every figure below is real; nothing is a projection.
- to the first theme
- 8.7s
- median, from raw feedback to a labeled, sentiment-scored theme, measured in production
- widget footprint
- 5.3KB
- framework-free, shadow-DOM isolated, one script tag
- hosted in the EU
- 100%
- data in Frankfurt, compute in the EU, PII redacted before storage
one real item, straight through the pipeline
“Love the new dashboard, but the export button is impossible to find”
Actual output from our production pipeline, theme label and emotions included. 8.7 seconds, end to end.
You are not short on feedback. You are short on a way to read it.
Sentriment reads all of it, prices what hurts, and hands you the words of the people who love you.
sentriment.com · EU-hosted · GDPR-first · PII redacted before storage
FAQ
Questions, answered.
PII (emails, names, phone and card numbers) is redacted before anything is stored; the raw text never touches the database. Everything is hosted in the EU (data in Frankfurt), encrypted in transit and at rest, never used to train third-party models, and deletable via API at any time. SOC 2 is on the roadmap; the GDPR architecture is live today.