Data Pipeline Monitoring

Your ETL job ran. It processed zero records. Nobody knew.

NotiLens monitors ETL jobs, batch syncs, data freshness, and row count anomalies — alerting your team the moment a data pipeline fails silently or produces suspicious output.

Start 7-Day Free Trial See All Solutions

Works with your data stack

Airflow

dbt

Fivetran

Airbyte

Prefect

Webhook

Zero-Record Alert — ETL

"nightly-user-sync completed. 0 rows processed."

Your ETL job ran on schedule, reported success, and wrote nothing. The source query returned empty. A schema change upstream silently broke the filter. Your analytics tables were stale for 18 hours before a data analyst noticed the dashboards hadn't moved.

nightly-user-sync · Zero-Record Alert · 0 rows · expected 12,000+

Silence Alert — batch sync

"orders-to-warehouse sync hasn't run in 4 hours."

Your batch sync is scheduled every hour. After four missed windows, NotiLens fired. The Fivetran connector had hit a rate limit, paused silently, and your warehouse was showing yesterday's order data to every team using it downstream.

orders-to-warehouse · Silence Alert · 0 runs in 4h · expected hourly

Row Count Anomaly — ingest

"events-ingest: 47,000 rows today. Baseline is 4,200."

An upstream bug started duplicating event records. Your pipeline ingested 11× the normal row count, passed validation, and wrote corrupt data to production tables. NotiLens caught the anomaly before your models ran on bad data.

events-ingest · Row Count Anomaly · 11× baseline

Freshness Alert — data warehouse

"customers table hasn't been updated in 6 hours."

Your customers table should refresh every two hours from your CRM sync. A silent authentication failure in the connector stopped all updates. Reports were running on stale data — and the business team had no idea until NotiLens fired a freshness alert.

customers-table · Freshness Alert · last updated 6h ago

Zero-Record Alert — dbt model

"revenue_daily model ran. Output: 0 rows."

A dbt model that aggregates daily revenue produced an empty table after a join condition broke silently. Finance reports showed $0 revenue for the day. The model "succeeded" in dbt — it just returned nothing. NotiLens caught it because the output row count dropped to zero.

revenue_daily · Zero-Record Alert · 0 rows · join condition broken

Silence Alert — pipeline

"product-events pipeline hasn't fired in 3 hours."

Your product analytics pipeline processes events every 30 minutes. After three hours of silence, NotiLens alerted. A Kafka consumer had stalled silently — events were queuing up upstream with no processing. The backlog had grown to 180,000 events before anyone knew.

product-events · Silence Alert · 0 runs in 3h · 180k events queued

The Problem

Data pipelines fail silently. Your monitoring doesn't see it.

ETL jobs report success even when they write zero rows. Batch syncs miss their window with no error. Dashboards go stale while teams make decisions on outdated data. The failure is invisible until the damage is done.

ETL job "succeeded" with zero rows

Your pipeline ran, completed without errors, and inserted nothing. A schema change upstream broke the source query silently. Standard monitoring saw a green checkmark — NotiLens would have seen zero rows where 12,000 were expected.

Batch sync missed its window

Your hourly sync didn't run. Then it missed again. After four hours, someone noticed the dashboards hadn't moved. The connector had silently paused — no error, no alert, just stale data flowing into every report downstream.

Duplicate rows wrote corrupt data

An upstream bug duplicated records before they reached your pipeline. Validation passed. Row count was high — but not flagged as anomalous. Your models ran on bad data for six hours before a data analyst spotted the totals looked wrong.

How NotiLens Helps

Pipeline alerting that catches what dbt tests miss

Send a webhook from your pipeline at the end of each run. NotiLens watches for silence, zero records, row count anomalies, and freshness violations — and alerts your team immediately.

Zero-record detection

Send your pipeline's output row count to NotiLens. When a run completes with zero rows where thousands are expected, you get an immediate alert — even if the job reported success.

Missed-run silence detection

Define how often your pipeline should run. If NotiLens doesn't receive an event within that window, it fires a silence alert — catching missed syncs before your data goes stale for hours.

Row count anomaly detection

NotiLens learns your pipeline's normal row count range. When a run produces 10× or 0.1× the expected volume, it fires — catching both duplicate ingestion and silent data loss before models run on bad input.

Data freshness monitoring

Send a heartbeat from your sync after each successful write. NotiLens alerts when a table hasn't been updated within its expected freshness window — so downstream consumers always know when data is stale.

Works with any pipeline tool

Airflow, dbt, Prefect, Fivetran, Airbyte, custom Python scripts — if it can fire a webhook, it works with NotiLens. No agents, no SDK required. One HTTP call at the end of each run.

Team alerting & ACK

Route data pipeline alerts to the right team members through the NotiLens mobile app. Critical alerts repeat until someone acknowledges them — so a zero-record ETL at 2am doesn't sit unnoticed until morning standup.

Real Alerts

What NotiLens sends your data team

Every alert includes the pipeline name, expected vs. actual values, and the exact timestamp — so you can diagnose and fix faster.

NotiLens Live Feed — Data Pipelines

nightly-user-sync: 0 rows processed — expected 12,000+

Topic: etl-runs • Zero-Record Alert • 2 min ago

Critical

orders-to-warehouse: no run in 4 hours — connector may be paused

Topic: batch-syncs • Silence Alert • 38 min ago

Critical

events-ingest: 47,000 rows — 11× normal baseline (4,200)

Topic: etl-runs • Row Count Anomaly • 1 hr ago

Warning

customers table: last updated 6h ago — freshness threshold exceeded

Topic: data-freshness • Freshness Alert • 2 hr ago

Warning

A pipeline that completed successfully is not necessarily a pipeline that did anything.

Zero rows, stale tables, missed windows — these are the data failures that slip past your orchestrator's green checkmarks. NotiLens is built to catch exactly what orchestration tools can't: silent, successful-looking failures.

FAQ

Common questions

Is this a replacement for dbt tests or Great Expectations? +

No — dbt tests and Great Expectations validate data quality within a run. NotiLens operates at the pipeline level: it detects when runs don't happen at all, produce zero output, or deviate significantly from their normal volume. They're complementary — your tests catch bad data inside a run; NotiLens catches the run not behaving as expected from the outside.

How does NotiLens detect zero-record ETL runs? +

At the end of each pipeline run, you send a webhook to NotiLens with the row count as a metric value. NotiLens learns your normal range over time and fires when a run reports zero rows — or any count that falls outside your expected baseline, even if the job exit code was success.

Which pipeline tools does NotiLens support? +

Any tool that can make an HTTP request works with NotiLens. This includes Airflow (via callbacks or sensors), dbt (via post-hook webhooks), Prefect (via automations), Fivetran (via connector webhooks), Airbyte, custom Python/Bash scripts, and anything else that runs as a job. No agents or proprietary SDKs required.

My orchestrator already sends failure emails. Why do I need NotiLens? +

Orchestrator failure emails only fire when a job exits with an error code. Zero-record runs, stale tables, and missed windows don't trigger errors — they look like successes. NotiLens catches the class of failure your orchestrator is blind to: the job that ran fine and did nothing useful.

Stop discovering data issues from your analysts.

No credit card required — 7-day free trial.

Start 7-Day Free Trial