Datadog Data Observability Pricing: What Jobs Monitoring, Database Monitoring and Data Streams Cost

Short answer: Datadog sells a product called Data Observability, and on the public pricing page it is one thing: Jobs Monitoring. The Clusters plan, covering Databricks and Apache Spark, starts at 0.05 dollars per host per hour billed annually, or 0.072 dollars per host per hour on demand. The Serverless plan for Databricks is priced per job per hour and carries no published figure, only Contact Us for Pricing. Database Monitoring is a separate product at 70 dollars per database host per month annually, or 84 dollars on demand. Data Streams Monitoring has no standalone meter at all: it ships inside APM Pro, at 35 dollars per host per month annually or 42 dollars on demand. Checked on datadoghq.com/pricing on September 4, 2026.

The figure most buyers are actually looking for is not on that page. Datadog acquired Metaplane in April 2025, and Metaplane is the product that does table-level data quality monitoring and column-level lineage. As of September 2026 the word Metaplane does not appear anywhere on Datadog's pricing page. It still sells from its own site, on its own tiers, metered per monitored table.

What Datadog charges on the data side, in one table

Every figure below was read off Datadog's own rendered pricing page. Where a cell says no published figure, that is because Datadog does not publish one, not because it was hard to find.

The Clusters plan covers Databricks and Spark jobs running on Kubernetes, Amazon EMR and Google Dataproc, and its stated jobs are detecting failing and long-running jobs, pinpointing root cause, and reducing cost by finding misallocated clusters and inefficient jobs. The Serverless plan covers Databricks serverless jobs and adds recommendations to optimize code and queries.

The meter is compute time, and that changes the whole calculation

This is the part worth slowing down for, because Datadog's data observability meter is unlike anything else in this market. Line the vendors up and ask one question of each: what has to happen for the invoice to go up?

Datadog is the only one of those whose data observability bill responds to how long your Spark clusters stay up. Add fifty checks to the tables you already watch and the Datadog number does not move at all. Leave a badly tuned job running four hours instead of forty minutes and it moves a lot, on every host in the cluster, every night.

Run the arithmetic on a realistic shape. Twenty hosts, jobs running six hours a day, thirty days a month, at the annual rate of 0.05 dollars per host per hour: 20 times 6 times 30 times 0.05 comes to 180 dollars a month. That is genuinely cheap for what it does. Now take the same twenty hosts on a cluster somebody left running continuously: 20 times 730 times 0.05 is 730 dollars a month, and on the on-demand rate of 0.072 it is 1,051 dollars. The observability coverage is identical in both cases. The only thing that changed is idle compute. That is why a per-host-per-hour meter belongs on the same dashboard as the rest of your infrastructure spend, and why teams running this meter usually want cloud and SaaS cost monitoring watching the trend rather than discovering it on the invoice.

What Jobs Monitoring does not cover

Jobs Monitoring watches jobs. It tells you a Spark job failed, ran long, or burned more cluster than it needed. Those are real problems and it is a good product for them. But the failures that reach a dashboard and a CFO are usually not job failures, and this is the gap to be clear-eyed about before you buy.

A job can succeed and still leave you with bad data. An upstream API returns an empty payload and the job loads zero rows, successfully. A source system changes a column from integer to string and the job coerces it, successfully. A batch replays and every order lands twice, successfully. A partition writes on time with last month's numbers in it, successfully. In every one of those cases Jobs Monitoring shows green, because the job did what it was told.

Catching them requires checks on the table itself: row counts against their own history, freshness against the load schedule, schema against yesterday's schema, and distributions against a learned baseline. That is table-level data quality monitoring, and it is the thing Datadog bought Metaplane to get. It is not what the Jobs Monitoring meter prices.

Where Metaplane sits after the acquisition

Datadog announced the Metaplane acquisition in April 2025 and said Metaplane would keep supporting existing and new customers as a rebranded Metaplane by Datadog offering. Seventeen months later that is still how it works. Metaplane sells from metaplane.dev with its own published tiers, and those tiers are metered per monitored table, not per host hour.

The practical consequence for anyone budgeting: a quote for Datadog data observability and a quote for Metaplane are quotes for two different products on two different meters, and you may well need both. Jobs Monitoring tells you the pipeline ran badly. Metaplane tells you the output is wrong. If your shortlist assumed one Datadog line item covers both, rebuild the model.

One detail from Metaplane's own pricing page is worth carrying into that conversation. Monitored tables scale generously across its tiers, from 10 on Free to 100 on Pro to unlimited on Enterprise. Custom SQL monitors, meaning the assertions you write yourself against your own business logic, scale from 3 to 5 to 10. The ceiling on your own rules stays in single digits at every tier. There is more detail on that in our breakdown of Metaplane pricing.

Is Datadog cheaper than a dedicated data observability platform?

For the job it does, usually yes, and that is a fair thing to say. A per-host-hour rate starting at 0.05 dollars is a small number, and if you already run Datadog for infrastructure then adding Jobs Monitoring is an incremental line rather than a new vendor, a new contract and a new procurement cycle. That consolidation argument is real and it wins deals.

The comparison stops being like-for-like the moment you include table-level checks. Then you are pricing Datadog plus Metaplane against a single platform that does both, and the meters no longer line up: host hours plus monitored tables versus one published monthly tier. It is also worth knowing what the alternative end of the market costs, since the enterprise listings are the only other public figures available. Monte Carlo lists a 12-month AWS Marketplace contract at 50,000 dollars metered per monitor, and Sifflet lists one at 48,000 dollars metered per asset. Two contracts four percent apart, measuring completely different things. Our data observability pricing comparison lays out all of the published figures side by side.

Which data warehouses does the Datadog data observability meter support?

The published Jobs Monitoring plans name Databricks and Apache Spark, with the Clusters plan covering Spark on Kubernetes, Amazon EMR and Google Dataproc, and the Serverless plan covering Databricks serverless. Snowflake, BigQuery and Redshift are not named in the Data Observability pricing plans. Database Monitoring is a separate product with its own per-database-host meter and a different purpose: it surfaces slow queries and correlates query performance with database infrastructure metrics, which is query performance work rather than data quality work.

If your estate is Snowflake, BigQuery or Redshift and your problem is knowing when a table is stale, short a million rows, or silently reschemaed, the Jobs Monitoring meter is not aimed at you. Warehouse-native freshness, volume, schema and distribution monitoring with alerts in Slack and PagerDuty is what our pipeline monitoring product does, read only, starting at 99 dollars a month with a 14-day trial and no card.

How to price this before you sign anything

Pull the four numbers that decide the bill, in this order. First, your average host count across the clusters you would instrument, taken over a full quarter so that month-end loads are inside the average rather than surprising you later. Second, average daily job runtime on those clusters, which is the multiplier most teams underestimate. Third, whether any of your Databricks footprint is serverless, because that half is quote-only and cannot be modeled from public information. Fourth, your database host count if Database Monitoring is in scope, at 70 dollars each per month.

Then ask the question the pricing page will not answer for you: which of the failures that actually hurt you last year would this have caught? Go back through the incidents. Count how many were a job that failed or ran long, and how many were a job that succeeded with wrong data in it. If the second pile is bigger, the Jobs Monitoring meter is priced for the wrong problem no matter how attractive the per-hour rate looks.

How much does Datadog Jobs Monitoring cost per hour?

The Clusters plan for Databricks and Apache Spark starts at 0.05 dollars per host per hour when billed annually, and 0.072 dollars per host per hour on demand. Both figures are from Datadog's public pricing page as of September 4, 2026. The Serverless plan for Databricks is metered per job per hour and Datadog publishes no rate for it, showing Contact Us for Pricing instead.

Does Datadog have data quality monitoring?

Not under the Data Observability product on its pricing page, which is Jobs Monitoring and meters pipeline compute. Table-level data quality monitoring came into Datadog through the Metaplane acquisition in April 2025, and Metaplane still sells separately from metaplane.dev on a per-monitored-table meter. A Datadog Jobs Monitoring subscription does not include it.

Is Data Streams Monitoring sold separately?

No. On the public pricing page Data Streams Monitoring is a feature of the APM Pro plan rather than a product with its own meter. APM Pro starts at 35 dollars per host per month billed annually, or 42 dollars per host per month on demand, and includes automatic dependency mapping of queues, consumers and producers, end-to-end pipeline latency and consumer lag metrics, and faulty queue detection. That is streaming infrastructure telemetry, not data quality.

What does Datadog Database Monitoring cost?

Database Monitoring starts at 70 dollars per database host per month billed annually, or 84 dollars per database host per month on demand. It surfaces slow performing queries, tracks normalized query performance trends from database-generated metrics, and correlates query performance with database infrastructure metrics. It is a database performance product and it does not check whether the data in those databases is correct.