The workflow

Four steps.
None of them take a quarter

Most analytics rollouts die in the modelling phase. Cadence is designed so the boring part is finished before lunch and the useful part starts the same day.

01Install

One script tag, one SDK call, or a warehouse connection. Events arrive within about five minutes.

02Name things once

Cadence proposes a schema from what it sees. You approve names, mark the sensitive ones, and move on.

03Explore

Funnels, retention, paths and segments — rebuilt in under a second, so you can keep asking.

04Automate

Save the metrics that matter. Cadence watches them and pings the channel that owns each one.

Step 01 · install

Five minutes, one snippet

Drop the tag in your layout, or install the package your framework already uses. Cadence starts collecting page views, clicks and session boundaries immediately — you don't have to instrument anything to see the first chart.

  • 3.1 kB gzipped, loaded async
  • Auto-capture on, opt-out per element
  • Backfill the last 90 days from your warehouse
Read the quickstart
Connected sourcesStreaming
Web SDK1.94M todayok
Mobile1.12M todayok
Stripe412k todayok
Snowflakebackfill 90d62%

Step 02 · model

The schema writes its own first draft

Cadence groups what it receives, spots the near-duplicates — signup_done, SignupComplete, signup-finished — and proposes one name for the set. You approve or rename, and every historical event is remapped, not just the new ones.

  • Merges apply retroactively to old data
  • PII fields flagged and masked on arrival
  • Every rename is logged with who and why
Suggested merges4 pending
signup_done+ 2 variants1.2M
checkout_started+ 3 variants860k
invite_sent+ 1 variant214k
board_created+ 2 variants96k

Step 03 · explore

Ask, look, ask again

Funnels rebuild while you're still moving the mouse. Add a breakdown, drop a step, switch the window from 30 to 90 days — none of it queues, so the shape of the answer stays in your head long enough to be useful.

  • Plain-English queries with visible SQL
  • Any chart becomes a saved metric in one click
  • Share a link that keeps the exact filter state
Signup → activationBreakdown: plan
Visited site128,400
Created account41,220
Activated18,940
Upgraded6,110

Step 04 · automate

Stop checking dashboards

A dashboard nobody opens is a dashboard nobody trusts. Cadence learns the normal shape of each saved metric — including its weekly rhythm — and only speaks up when a cohort genuinely breaks its own pattern.

  • Seasonality-aware, not fixed thresholds
  • Routed per metric owner, not one firehose channel
  • Monday digest with what actually moved
Recent alerts2 active
Checkout drop-offJul 21 cohort+12.4%
Mobile activationAndroid only−4.1%
Trial → paidEU region+6.8%
Weekly digest#growthMon 9:00

What the first month usually looks like

Averaged across the last 300 workspaces that onboarded without a solutions engineer.

Day 1

Install and first chart. The snippet goes in, auto-capture starts, and someone shares a link to a page-view breakdown before the end of the day.

Day 2–3

Naming. You accept most of the proposed schema, rename a handful of events, and mark two or three properties as sensitive.

Week 1

The activation funnel exists. It has an owner, a definition everyone has read, and a number that stops being argued about.

Week 2

Retention baselines. Weekly cohorts go back as far as your warehouse backfill reached, so you're comparing against real history, not the day you installed.

Week 3–4

Alerts and the Monday digest. People stop opening dashboards out of anxiety, because the tool tells them when there's something to see.

0msp95 ingest latency, event to queryable
0sp95 query time across 8B rows
0native sources and sinks
0%ingestion pipeline uptime

Questions people ask on the first call

Do we need a data engineer to set this up?

No. The install is a script tag or a package, and the schema step is a review screen, not a modelling project. Teams with a warehouse usually connect it in the first week as well, but nothing depends on that.

What happens to the data we already have?

If it lives in a warehouse, Cadence backfills up to 24 months on request so your retention curves start with real history. If it lives in another analytics tool, we import from Amplitude, Mixpanel and Heap exports directly.

Can we keep using our own SQL?

Yes. Every chart exposes the query it ran, you can edit it, and you can query the underlying tables through the read-only Postgres-compatible endpoint on Business and above.

How do you handle sensitive properties?

Anything you mark sensitive is hashed or dropped at the edge, before it is written to storage. Session replay masks all input values by default — you opt specific fields in, not out.

What does migration off Cadence look like?

One button produces a full export of raw events in Parquet plus your metric definitions as JSON. We don't hold the export hostage behind a support ticket, and it works on every plan including Free.

Is there a limit on seats?

Only on Free, which is capped at three. Every paid plan is unlimited — you're billed on events, so inviting the whole company costs nothing and is usually the point.

Get started

Ship on Monday.
Understand it by Tuesday.

Free for 30 days on the full product. Five-minute install, no card, no onboarding call unless you want one.

  • SOC 2 Type II
  • EU & US data residency
  • Cancel in one click