Features — Cadence

Ask in plain English

Type the question the way you'd say it in standup. Cadence writes the query, picks the chart, and shows the SQL underneath so you can check its work.

Median answer: 0.4s

Funnels & paths

Build a funnel in four clicks, then break it down by plan, device, country or any property you've ever sent. Reordering steps re-runs in under a second.

Unlimited steps · 90-day windows

Retention & cohorts

Weekly, monthly or rolling retention on any event pair. Save a cohort once and it stays live — new users flow in as they qualify.

Live cohorts · no re-runs

Live event pipeline

Web, mobile, billing and your warehouse land in one schema. Cadence de-dupes on arrival, backfills late events and keeps property names consistent.

p95 ingest latency 840ms

One shared metric library

Define activation once, with an owner and a description. Every dashboard, alert and export reads that definition — so reviews stop turning into audits.

Versioned · owner per metric

Session replay

Jump from a drop-off step straight into the sessions behind it. Inputs, tokens and anything you mark sensitive are masked before they leave the browser.

Masked by default

Alerts that find you

Cadence watches every saved metric for a break in its own pattern — not a fixed threshold you'll forget to update — and posts to the channel that owns it.

Slack · Teams · email · webhook

Experiments

Point Cadence at your flag provider and it reads exposure events automatically. Sequential testing, so you can look at the result early without lying to yourself.

LaunchDarkly · Statsig · custom

Two-way warehouse sync

Read modelled tables straight out of Snowflake or BigQuery, and push Cadence cohorts back down so the rest of your stack can use them.

Snowflake · BigQuery · Redshift

Question to answer

The query is the easy part. Trusting it is the hard part.

Every answer Cadence gives you comes with the definition it used, the rows it counted and the person who owns that metric. Disagree with the number and you can see exactly where it came from instead of arguing from memory.

Generated SQL is visible on every chart
Definitions are versioned, with a changelog
Every metric has a named owner

Activation rateOwned by Product

Definitionimported in session 1v4

Reads it18 dashboards6 alerts

Speed

Fast enough to keep thinking

Cadence stores events in a column-oriented engine built for this one job. A funnel across eight billion rows rebuilds in about a second, which means you can change the question mid-thought instead of scheduling a re-run and losing the thread.

p95 query time 1.4s on 8B rows
No pre-aggregation to maintain
Breakdowns don't cost extra time

Query logLive

Funnel · signup → paid8.2B rows1.1s

Retention · weekly4.6B rows0.8s

Paths · first session2.1B rows0.6s

Breakdown · by plan8.2B rows1.3s

How it compares

The honest version. Cadence is not a warehouse and not a BI tool — it does one layer, properly.

	Cadence	Spreadsheets + BI	Legacy product analytics
Time to first real insight	Same afternoon	Two to six weeks	One to two weeks
Who can answer a question	Anyone on the team	Whoever writes SQL	Trained analysts
Billions of rows, ad hoc		Scheduled runs	Sampled above 10M
Single metric definition		Per workbook	Per dashboard
Anomaly alerting	Built in, per metric	Manual thresholds	Add-on
Session replay in the same tool		—	Separate product
Two-way warehouse sync		Read only	Export only
Pricing model	Events, not seats	Per seat	Per tracked user
EU data residency		Depends	Enterprise only