Data pipeline observability

Know your data is late before your CFO does.

Tidewater watches your Airflow, dbt, Spark and Kafka pipelines, flags data that is late, short or changed, and shows which dashboards it affects and who owns them.

Request a demo See how it works

Free for up to 10 pipelines. Median time to detect late data: 4 minutes.

payments_ingest

RunningOn time

Airflow, hourly

orders_daily

Due 06:00Late by 38 min

dbt, due 06:00

inventory_snapshot

On time

Spark, every 6 h

events_stream

Lag 2 s

Kafka, continuous

Impact of orders_daily 5 assets

Runs, last 6 hours US Eastern

payments_ingest

orders_daily

inventory_snapshot

events_stream

due 06:00

now 06:38

02:0004:0006:0008:00

Trusted by data teams at 140 companies, from freight to credit to grocery.

Northfold Freight
Kestrel Credit
Lumen Grocer
Aldermoor
Tessaly
Corvane

What it does

Three jobs, done before anyone asks.

Finding late data, finding what it affects, and telling the right person. Each one is a screen you will actually use.

01 / 03

Monitoring

Say when data should arrive. Tidewater learns the rest.

Write the deadline in plain words, such as "updated by 06:00 on weekdays". Tidewater learns the usual arrival window, the weekday pattern and the normal row count from your history, then tells you when a table misses.

Freshness, volume and schema checks on every pipeline
Thresholds learned from your own history, adjustable by hand
Catches the job that succeeds on yesterday's data

See monitoring

Arrival time Late today

Usually lands 05:00 to 05:37

Your deadline 06:00

Today 06:38

Tidewater learned the usual window from 13 days of history. You only choose the deadline.

02 / 03

Lineage

See what a late table touches before you message anyone.

Tidewater reads your SQL and your dbt project to draw what feeds what, from source tables to the dashboards people open. When a table is late, the graph shows exactly which assets are affected and who owns each one.

Built from your own queries, so it is never out of date
Upstream to find the cause, downstream to size the impact
Owners and last-viewed times on every node

See lineage

03 / 03

Alerts

Send the alert to the person who owns the table.

Alerts go to the owning team's channel with the impact already attached. Rules decide who is paged, who waits for the morning digest, and when a quiet hour applies.

Routing by owner, severity and time of day
Slack, Teams, PagerDuty, Opsgenie and webhooks
One digest in the morning instead of eleven pings overnight

See alerts

Alert

orders_daily is 38 minutes late 06:38

Expected by 06:00 Eastern. Last run finished yesterday at 05:21.

Rules, first match wins

If table owner is finance-data Notify #finance-data-alerts

If severity is high and 3 or more assets affected Page the on-call

If between 22:00 and 07:00 and severity is low Hold for the morning digest

Delivered to

#finance-data-alerts Slack

orders_daily is 38 minutes late.
5 assets affected, owner finance-data.

Acknowledge Open lineage

Data on-call PagerDuty

Morning digest Email

How it works

From connected to useful in an afternoon.

Tidewater sits beside your stack and reads metadata. It does not sit in the path of your data and it never reads table contents.

Connect

Give Tidewater read access to your orchestrator and warehouse. It reads run events and metadata only.
About 15 minutes
Learn

Tidewater studies your run history to learn when each table normally lands and how big it is.
Two weeks to full accuracy
Detect

Every minute it checks for late, missing, short or changed data, including jobs that report success.
Median detection 4 minutes
Explain

Each alert names the owner, the cause it found upstream and the dashboards downstream.
Impact in every alert

Integrations

Works with the tools you already run.

Twenty-two integrations across orchestration, transformation, compute, streaming, warehouses, BI and alerting.

Apache Airflow Orchestration
Dagster Orchestration
Prefect Orchestration Beta
dbt Cloud Transformation
dbt Core Transformation
Apache Spark Compute
Databricks Compute
Amazon Kinesis Streaming Beta
Apache Kafka Streaming
Amazon Redshift Warehouse
ClickHouse Warehouse Beta
Google BigQuery Warehouse
PostgreSQL Warehouse
Snowflake Warehouse
Looker BI
Metabase BI Beta
Tableau BI
Microsoft Teams Alerting
Opsgenie Alerting
PagerDuty Alerting
Slack Alerting
Webhooks Alerting

Northfold Freight

We used to learn that the shipment tables were stale from a customer email. Now the first message is in our own channel, and it already names the dashboards it affects.

Marisol Vance, Director of Data Platform, Northfold Freight

Read the Northfold Freight story

6 min Median time to detect late data
71% Fewer late data incidents reaching customers
4 Hours a week saved on triage

Pricing

Priced by pipeline, with unlimited seats.

Start free with 10 pipelines. Pay for the pipelines you monitor, not for the people who read the alerts.

Free

Try Tidewater on a real pipeline. No card, no time limit.

$0 per month, includes 10 pipelines

Monitored pipelines: Up to 10
History retention: 7 days
Shortest check interval: 15 minutes
Alert channels: 1 (Slack or e-mail)

Start with the quickstart

Team

Most teams start here

For a data team that owns its pipelines and wants to hear about problems first.

$149 per month, includes 25 pipelines

Monitored pipelines: 25 included, up to 250
History retention: 30 days
Shortest check interval: 5 minutes
Alert channels: Unlimited

Request a demo

Business

For platforms that serve many teams and need column lineage, SSO and a firm response time.

$749 per month, includes 150 pipelines

Monitored pipelines: 150 included, up to 2,000
History retention: 13 months
Shortest check interval: 1 minute
Alert channels: Unlimited

Request a demo

Estimate your price

Questions

Before you ask.

How is Tidewater different from the tests I already run in dbt or Airflow?

Tests tell you that a job failed or a rule broke. Tidewater also tells you when data is late while every job reports success, which is the case that reaches dashboards unnoticed. It then shows which downstream tables, models and dashboards are affected, and who owns them.

How long does setup take?

Most teams connect an orchestrator and a warehouse in under an hour and see their first lineage graph the same day. Learned thresholds need about two weeks of history before they start alerting, and you can set fixed thresholds on day one.

Will Tidewater slow down my pipelines or my warehouse?

No. Tidewater reads events and metadata that your systems already produce. Warehouse queries use your own read-only role, run on a schedule you control, and are tagged so you can see their cost in your own query history.

Can I use Tidewater without an orchestrator?

Yes. You can monitor tables directly in the warehouse, or send run events yourself with one HTTPS call from any script. The quickstart shows both.

See your own pipelines in Tidewater

Book a 30 minute walkthrough. We connect a sample of your setup and show you what it would have caught last month.

Request a demo Read the docs