Nimbus turns raw product events into funnels, retention cohorts and anomaly alerts that a designer, a founder and a data engineer can all read from the same screen — without one of them writing SQL for the other two.
No card required · 2M events/month free · 4-line install
Events ingested today
▲ 6.2% vs. same time yesterday
Type the question. Nimbus builds the funnel, marks the step that is bleeding and shows you which segment it is bleeding from — before you have decided which chart you wanted.
Any step over a 40% drop is flagged automatically and broken down by device, plan and source.
Nimbus watches every funnel step, every cohort and every event volume against its own seasonal baseline. When something moves past what the day of the week explains, you get one sentence in Slack — the metric, the size of the move and the segment it came from.
Ask in a sentence and you get the chart. Open the query drawer and you get the exact SQL that produced it, editable, versioned and copyable. Nothing is hidden behind a black box, which is what makes the answer arguable — and therefore useful.
Events arrive misspelled, duplicated and half-typed. Nimbus proposes merges for near-identical names, keeps a reviewable history of every rename and never silently drops a property, so the taxonomy stays honest after the third engineer joins.
Events land in a columnar store partitioned by day and sorted by user, with rollups maintained on write. A 12-month retention cohort over a billion events returns in about 340 ms, which is the difference between exploring a question and filing a ticket about it.
One script tag or one npm package. It is 4.1 kB gzipped, batches on an idle callback and never blocks a paint. Server-side and mobile SDKs share the same event contract.
Point Nimbus at Postgres, BigQuery, Snowflake or a CSV drop and it replays your history into the same schema. Most teams import 12 months in an afternoon and open a real cohort table the same day.
No dashboard to assemble before you get value. Type a sentence, get the chart, pin it if it turns out to matter. Alerts are created from any pinned chart with one click.
// 1. install
npm i @nimbus/analytics
// 2. identify + track
import nimbus from '@nimbus/analytics'
nimbus.init({ key: 'pk_live_9f2c…' })
nimbus.identify(user.id, {
plan: user.plan,
company: user.orgId,
})
nimbus.track('workspace_created', {
source: 'onboarding',
template: 'blank',
seats: 3,
})4.1 kB gzipped · no cookies by default · GDPR mode on for EU traffic
Weekly cohorts, twelve columns deep, rebuilt on every query rather than cached overnight. When the March release moved week-4 retention, you can see the row it moved.
| Cohort | Size | W0 | W1 | W2 | W3 | W4 | W5 | W6 | W7 |
|---|
“Our onboarding lost half its users at one step for eleven months. Nimbus put it on screen in the first session, in a sentence, with the browser attached.”
“The SQL drawer is why we trust it. Every number can be opened, argued with and corrected — nothing has to be taken on faith.”
“Ad-hoc requests to my team dropped by about two thirds. Not because people ask less, but because they can answer themselves.”
Charging per seat teaches a company to hide its data from itself. Everyone gets a login on every tier, including the free one.
2M events, 6-month history, unlimited seats, funnels and cohorts, community support.
Start50M events, 3-year history, anomaly alerts, SQL drawer, warehouse sync, SSO, 4-hour support.
Start 14-day trialUnlimited history, EU or US data residency, audit export, 99.99% SLA, named analytics engineer.
Talk to usNo. Nimbus reads the event names you already send from Segment, RudderStack, PostHog or your own pipeline, and maps them on the way in. Teams typically keep their existing SDK for a quarter and run both, then delete the old one when the numbers agree.
Events are stored columnar, partitioned by day and sorted by user id, with funnel and retention rollups maintained on write rather than computed on read. A 12-month cohort touches a few hundred megabytes rather than the whole table.
It writes SQL. The generated query is shown in a drawer next to every chart — you can read it, edit it, save it as a template and hand it to someone who prefers writing queries by hand. If the sentence was ambiguous, Nimbus says which reading it took.
Frankfurt by default for EU accounts, with US and Singapore available. Region is set per workspace at creation, data never leaves it, and the choice is visible in the workspace settings rather than buried in a contract.
Yes — continuous sync to Snowflake, BigQuery or S3 in Parquet, plus a full export API. Your raw events are yours, in an open format, at all times.
Drop in four lines, backfill your history and open the first cohort table this afternoon.