Sonder
Sign in Join the beta

Know what to improve next.
Then know it worked.

Sonder continuously analyzes the product behavior you already collect to find where users struggle, what is causing it, and what to change—then watches every release to prove the product got better.

Show me what to improve See one improvement end to end
Continuously improve Onboarding Core workflows Reliability · from data you already collect
RESEARCH RUN 009 Example data
Where the product gets stuck

2,481 sessions analyzed

100%
intentOrganize a project before export×44
frictionTag selection never lands×63
patternSame failure across sessions0.91
PRODUCT ISSUE · HIGH CONFIDENCE

The menu unmounts before the selection commits.

See recommended change →
AFTER RELEASE 4.2→0.6 failed attempts / session
Live product signal
The problem

Your replay library is full of answers.
Sonder puts them to work.

Every abandoned task, repeated click, and unfinished workflow is a clue to a better product. Sonder reads them all, continuously, so the most important improvements rise to the top.

What continuous reading feels like 2,481 sessions → 3 issues
Tag selection never lands44 users · introduced in v2.3 Worsening
New teams miss first import31 trials · first session Watching
Export fails without feedback9 users · fixed in v2.4.1 Resolved
Without Sonder Hours of footage. The answers stay buried.
With Sonder A ranked plan to make the product better.

From a hidden problem to
a measurable improvement.

Sonder catches the moments that make your product harder to use before they become support tickets or roadmap debates, shows your team why they happen and what to change, then watches new sessions to prove the experience improved.

ISSUE UX–07 Example data Resolved after 18 clean sessions
01ObserveOne stuck session
Session 9fa1 · Tuesday 09:38
view/projects/atlas
click“Tag dropdown”
select“Q3 launch”
rageclick“Tag dropdown” ×4
leave/projects/atlas
What Sonder understood

This user is organizing a project before export. The menu keeps closing before the selection lands, so they repeat the same action and leave.

You now know it is a product problem: forty-three other people hit the same failure. Filed into UX–07 · Tag selection never lands
  1. 02DiagnoseLikely cause
    UX–07 Worsening
    Tag selection never lands The menu unmounts before the selection commits.
    Affected
    44 users
    Confidence
    91%

    Know the scale, likely cause, and confidence without watching sixty-three replays.

  2. 03ReviewScoped proposal
    TagPicker.tsx8 lines
    1. 12− closeMenu()
    2. 13selection.commit()
    3. 14+ await settleSelection()
    4. 15+ closeMenu()

    Your team gets a small, evidence-backed change to review, not another research project.

  3. 04VerifyBehavior changed
    v2.4.1 · AFTER SHIP Resolved
    Failed attempts / session−86%
    4.2 → 0.6
    Clean sessions
    18
    Target
    < 1.0

    Know the change worked, and if the problem ever comes back.

See how Sonder improves your product
The resolution protocol

See which releases actually made the product better.

Sonder watches real behavior after every ship, so you can see what improved, what came back, and where to act next without building another dashboard.

ShipLearnImprove
  1. 01 ResolvedBehavior improved, and held Closes
  2. 02 ReturnedThe pattern came back, issue reopens Reopens
  3. 03 WatchingNot enough signal yet, stays open Stays open
Data posture

Watch the product.
Not the person.

  • No pixels
  • No keystrokes
  • No replay video

Sonder reads the structure around friction: roles, labels, routes, events, and a small masked state snapshot. If it wouldn’t belong in a bug report, Sonder never collects it.

route/projects/:id rolebutton “Add tag” menu_openfalse typed_value[never collected] session_replay.mp4not collected
See exactly what Sonder collects
Setup

Put the data you already collect to work.

Connect PostHog or Amplitude in about three minutes. Sonder starts finding the highest-leverage ways to improve your product; add richer context only when it earns its place.

Read-only import3-minute setup
PostHogorAmplitude

Unlock the insights already there.

Choose a project and let Sonder turn the events you already trust into a ranked improvement plan.

  • No tracking-plan rewrite
  • No replay export
  • Read-only access
Read the import guide
Sonder SDKAdd context later
terminal
npx sonder-init <write_key>

Go deeper where it matters.

Add roles and masked state to understand your most important workflows in more detail.

  • Typed, inspectable events
  • Masked by default
  • Works with your current stack
Open the quickstart

A few things worth knowing.

For implementation details, data handling, and the SDK, read the docs

Does Sonder record my users?
No. It reads routes, roles, labels, events, and masked state, not pixels, keystrokes, or replay video.
Can I start without code?
Yes. Connect PostHog or Amplitude read-only and get a ranked improvement plan from the data you already have. Add the SDK only where richer structure is useful.
Does Sonder ship fixes?
You stay in control. Sonder proposes a scoped change with evidence. Your team reviews and ships it; Sonder confirms whether the experience improved.
What does the beta cost?
Founding beta workspaces are free. Later plans are based on tracked issues, never session volume.

Ship at AI speed.

Know what to improve next. Then know it worked.

Show me what to improve