Metric alerting for analysts. No engineers required.

Describe any metric in plain English. Lighthouse writes the SQL, learns your baseline, and sends a Slack alert the moment something meaningful changes.

Start for free

Free forever tier
Read-only access
No credit card

Lighthouse is metric alerting for analysts, with no engineers required. Describe a metric in plain English and Lighthouse writes the SQL, builds a baseline from the same weekday and hour over the last 4 weeks, and sends a Slack alert with value, baseline, threshold and segment when something meaningful changes.

Transactions

812-34%

Transactions triggered

Value812

Baseline1,340

Most teams find out about metric drops in the weekly review.

By then, it's been broken for days. Your options today aren't great.

Today

Option ATakes weeks. Breaks silently.
Build custom alerting
A few weeks of engineering time to build Cloud Scheduler + SQL + Slack. It works until the schema changes and breaks silently. Nobody notices until Monday's stand-up.
Option BNoisy. Tuned out.
Use BI tool alerts
Static percentage thresholds on a Looker tile. Fires every weekend, every holiday, every campaign spike. Your team stops reading Slack alerts entirely.
Option CToo slow. Too late.
Hope someone notices
The de facto choice at most companies. Someone catches a drop in a dashboard during a review meeting, three days after it happened.

With Lighthouse

Describe it. Watch it. Get pinged.

The full stack from metric definition to Slack alert, in minutes, not sprints.

1
Describe in plain English
“Transactions in the last 3 hours, drop > 30%”
2
Lighthouse generates SQL
Complex BigQuery / Snowflake queries with baselines, window functions, and filters.
3
Baseline is learned
Lighthouse learns seasonal patterns and normal ranges for your specific metric.
4
Slack alert fires
When the metric moves, your team gets context: value, baseline, threshold, segment.

Every feature your team needs to stop missing things.

Plain English metric creation

Describe what you want to monitor in a sentence. Lighthouse writes the SQL, configures compare periods, sets severity, and picks up your alert delivery.

Works on any SQL warehouse

Snowflake, BigQuery, Postgres, or SQL Server. Connect with read-only credentials and point Lighthouse at any table, view, or materialized view.

Adaptive anomaly detection

No static percentage thresholds. Lighthouse learns day-of-week patterns and seasonal baselines, and fires only when something genuinely moved.

Severity levels (S1 / S2)

Tag metrics as S1 (critical, alert immediately) or S2 (notable, batch digest). Route high-priority alerts to #data-oncall and lower ones to #data-digest.

Segment-aware alerts

Monitor the same metric sliced by region, user type, platform, or any column in your schema. Get alerted only when a specific segment breaks.

Snooze & deduplication

Consecutive-trigger logic prevents alert storms. If a metric stays in breach, Lighthouse deduplicates and only re-alerts when the situation changes meaningfully.

Built for every person on the data team.

Data analysts

Set up monitoring on the metrics you own without filing a ticket or writing a single line of SQL. Know before your stakeholders do.

Analytics leads & heads of data

Give your whole team a monitoring layer above the warehouse. Stop being the person people Slack when something looks off. Lighthouse does that automatically.

Product managers

Monitor your own product KPIs without waiting for a data request. Get pinged in Slack the moment a metric you care about moves outside its normal range.

Our PM team can now track their own KPIs without filing a ticket. Lighthouse is the first monitoring tool that non-technical people actually use.

Director of Product

A fast-growing e-commerce platform

We used to get 200+ alerts a week and had tuned most of them out. Lighthouse got that down to under 20, and every one of those 20 was real.

Head of Data

A top-10 mobile gaming studio

Common questions

What's wrong with static threshold alerts?

Static thresholds (e.g. 'alert when transactions drop more than 20%') fire every weekend, every holiday, and every time you run a campaign that inflates a baseline. Teams tune them out. The one real drop gets buried. Lighthouse learns your patterns and only alerts when something is genuinely anomalous.

How does the AI learn my baseline?

Lighthouse looks at historical data for the same time window (same day of week, same hour) across the last 4 weeks. That becomes the baseline. If Tuesday at 10am is always lower than Monday, Lighthouse knows that and won't alert on it.

Can I write my own SQL instead of using plain English?

Yes. You can use plain English (recommended for most metrics), paste your own SQL, or edit the AI-generated SQL. Lighthouse is flexible: use whatever fits your team's workflow.

What warehouses does Lighthouse support?

Snowflake, BigQuery, Postgres, and SQL Server. Connect with read-only credentials. We never write to your warehouse.

How many Slack alerts will we actually get?

Much fewer than you'd expect. Consecutive-trigger logic means Lighthouse waits for 2 back-to-back anomalies before firing, snoozing is on by default, and severity levels let you route lower-priority signals to a digest. Most teams go from 200+ alerts/week to under 20, and every one matters.

Does Lighthouse replace our BI tool?

No, and it shouldn't. Lighthouse is proactive alerting; your BI tool is reactive exploration. They complement each other. Lighthouse tells you something moved; your dashboard tells you why.

Stop finding out in the weekly review.

Set up your first metric alert in 10 minutes. Free forever for small teams.