23 data sources
23 data sources, and alerts that explain themselves
Lighthouse now watches metrics wherever they live: warehouses, databases, product analytics, BI tools, Stripe, HubSpot and Google Sheets. Alerts arrive with the context to act on them, and the Overview opens on what needs you.
Data sources
23 sources, every one read-only
New this month: Databricks, a native Amazon Redshift connector, MySQL, Oracle Database, ClickHouse, Trino, Amazon Athena, Amplitude, PostHog, Google Analytics 4, Tableau, Metabase, Grafana, Stripe, HubSpot and Google Sheets. Looker is out of early access and generally available. Every connection only reads.
- WarehouseSnowflake, BigQuery, Databricks, Amazon Redshift
- DatabasePostgres, SQL Server, MySQL, Oracle Database, ClickHouse
- Query engineTrino, Amazon Athena
- AnalyticsMixpanel, Amplitude, PostHog, Google Analytics 4
- BILooker, Tableau, Metabase
- MonitoringGrafana
- PaymentsStripe
- CRMHubSpot
- SpreadsheetGoogle Sheets
- HTTP APIAPI Connector, for any REST API
Flavors in the picker
- PostgresSupabase, Neon, Aurora PostgreSQL, AlloyDB, Timescale
- SQL ServerAzure SQL
Alerts
Alerts that carry their own context
An alert should answer the first three questions before anyone opens a dashboard: what broke, how bad, and who owns it. Read the alerts docs.
- Which rule matched, in plain words. The alert says what crossed the line and by how much, with how long it has lasted and the previous value.
- A 30-day chart in Slack. Single-segment alerts attach the segment's 30-day history as an image, so the trend is right there in the channel.
- Threads, not floods. Repeats and escalations go as replies under the first alert message instead of new posts.
- A message when it recovers. A metric can post when its alert recovers, and the original alert is marked recovered.
- The right people, by severity. Mention the metric owner, or a person or group once an alert reaches a severity you choose. Severity can be fixed, or scale with how far the value moves.
- A “What to do” note. Write the runbook step on the metric and it shows up in every alert it sends.
- History at a glance. An alert says when this segment last alerted and how that one was handled. Aggregated alerts list the most severe segments first.
- A daily digest and a weekly summary. One daily message with the alerts still open, and one weekly summary of runs, failed runs and metric changes.
- Quieter system notices. One notice when a data source becomes invalid, a choice between “A metric stops” and “Every failed run”, and each member picks how they get their own notices.
Microsoft Teams is a destination too: alerts arrive as interactive cards with the segment's history as a native chart.
Snowflake kit
31 Snowflake metrics, and a savings report
The Snowflake kit now covers cost, performance and failures in about two minutes, and tells you what you could cut. See the kit.
- 31 metrics in six groups. The original 13 monitoring metrics plus five cost groups: spend trends, warehouses, people and jobs, AI, and storage and indexes. Pick the groups at install.
- Sized to your account. The kit suggests small, medium or large from your last 30 days of credits, so a metric only alerts once it's worth your time.
- A savings report in dollars. Each finding comes with its evidence, the fix, the trade-off, and a realistic saving per month and per year. Run it from the Savings page, at the end of the install, or by asking Claude.
- Every fix watched. Each finding gets a metric that stays quiet while the fix holds and alerts if the waste comes back. A new report lands on the 1st of each month.
- Shareable. Admins and the kit owner can share a read-only link with user names hidden.
Overview, Metrics Manager and Audits
What needs you, first
- Needs attention, first. The Overview opens on a queue of what needs you, with failed metrics grouped by the cause they share, so one broken connection reads as one problem.
- Open alerts as New, Ongoing and Stale. Grouped by metric, beside system health and recent activity.
- Metrics Manager tabs. Live, Paused and Draft, with the metrics that are alerting right now marked and cron schedules described in words.
- Audits funnel cards. See how alerts were handled, as matches or triggers and as counts or rates.
- Kits where you need them. Connect a data source and Lighthouse offers the matching kit right away, also from the create menu and empty states. Organizations can publish their own kits too.
Also shipped
Roles, schedules and pricing
- Owner and Viewer roles. A new Owner role for the person who holds the organization, and a read-only Viewer role that can't act from Slack alert buttons either. Roles and permissions.
- Weekly and monthly schedules. Native weekly and monthly repeats, and a Today calculation window next to Yesterday.
- Business metric to SQL. Convert a form-built metric to SQL from the SQL view, with its thresholds, segments and alert variables carried over.
- API key expiry reminders. A reminder a week and a day before a key expires.
- Flat organization pricing. Plans are priced per organization, not per seat, billed monthly or annually. See pricing.
