At a glance
| Supported data platforms | Data Catalog tables on S3, Redshift, JDBC, Iceberg, Hudi, Delta Lake, S3 Tables and Athena views; any Glue ETL source in jobs |
| Anomaly detection | Yes, ML-based on collected statistics, in ETL jobs and (since July 2026) for Data Catalog tables |
| Freshness and volume monitoring | Through rules and analyzers (for example RowCount, DataFreshness, FileFreshness), not automatic |
| Schema checks | SchemaMatch, ColumnExists, ColumnCount and ColumnDataType rules; not an automatic schema change monitor |
| Validation rules | DQDL, more than 25 rule types including CustomSQL, ReferentialIntegrity and DatasetMatch; composite rules; dynamic thresholds in ETL |
| Lineage and root cause | Not documented; row-level failure identification in ETL jobs |
| Alerting | Amazon EventBridge and CloudWatch |
| Deployment | Serverless on AWS only |
| Licensing | Pay as you go in Glue DPU-hours; no licence fee |
AWS Glue Data Quality suits teams whose data lake and pipelines already run on AWS Glue and the Glue Data Catalog: rules sit next to the tables and jobs, results go to CloudWatch, EventBridge and S3, and you pay only for DPU time. It is AWS only and rule-centric, so teams wanting cross-platform observability with lineage-based root cause need another tool.
Overview
AWS Glue Data Quality is part of AWS Glue, not a separate product. AWS says it is built on the open source Deequ library and uses DQDL, a rules language with more than 25 built-in rule types such as completeness, uniqueness, referential integrity, data set comparison and file freshness. Each evaluation produces a data quality score, the percentage of rules that pass. Statistics gathered by rules and analyzers over time feed ML-based anomaly detection and, in Glue ETL jobs, dynamic rules such as RowCount > avg(last(10)).
There are two entry points. Data quality for the Data Catalog evaluates tables (S3, Redshift, JDBC, Iceberg, Hudi, Delta Lake, S3 Tables and Athena views) at rest, with rule recommendations and scheduling. Data quality in Glue ETL jobs checks data in transit, can identify and quarantine the failing records, and supports Flex and auto scaling. Recent additions include writing results to Iceberg tables in the Data Catalog and anomaly detection for cataloged tables (July 2026), and Bedrock-based ADVANCED rule recommendations and Athena view support (September 2026).
Key features
- DQDL rules language with more than 25 rule types, composite rules, WHERE filters and constants (labels in Glue ETL only)
- Rule recommendations for Data Catalog tables, with an ADVANCED mode that uses Amazon Bedrock (September 2026)
- ML-based anomaly detection, plus dynamic rules that compare metrics with their own history (dynamic rules work in Glue ETL only)
- Data quality checks in Glue ETL jobs that identify, filter and quarantine failing records
- Data quality scores with results sent to Amazon CloudWatch, Amazon EventBridge and Amazon S3, or Iceberg tables in the Data Catalog
- Incremental evaluation with pushdown predicates (Data Catalog) or job bookmarks (ETL)
- File-level rules such as FileFreshness, FileSize, FileUniqueness and FileMatch
Integrations and ecosystem
Pros and cons
Pros
- No infrastructure and no licence: billed in DPU time like the rest of AWS Glue
- Rules live next to Data Catalog tables and Glue jobs, with CloudFormation support
- Can stop bad records in ETL jobs before they reach the data lake
- Rules language based on open source Deequ, which reduces lock-in
Cons
- AWS only: no coverage of warehouses or pipelines outside AWS Glue sources
- Rules cannot evaluate nested or list-type data without flattening first
- Some features differ by entry point (recommendations only for the Data Catalog, failing-record identification only in ETL jobs)
- Ruleset limits of 2,000 rules and 65 KB, and 100,000 stored statistics per account
Pricing
AWS Glue Data Quality has no licence fee and is billed in AWS Glue DPU-hours. The AWS Glue pricing page uses USD 0.44 per DPU-hour (USD 0.29 for Flex) in its examples. Rule recommendation and evaluation tasks on Data Catalog tables need at least 2 DPUs with a 1-minute minimum billing duration (the page's examples: 5 DPUs for 10 minutes of recommendations is about USD 0.37, 5 DPUs for 20 minutes of evaluation about USD 0.73). Checks inside ETL jobs are billed as part of the job's extra run time or DPUs. Anomaly detection on Data Catalog tables costs 1 DPU per statistic for the detection time (10 to 20 seconds on average, 1-second minimum), and retraining also costs 1 DPU per statistic; anomaly detection in ETL jobs has no extra charge. Storing statistics is free (limit 100,000 per account, kept up to two years). Amazon S3 and Data Catalog charges apply separately, and rates vary by region.
Pricing checked on the vendor's official page on 7 November 2026. Prices change; confirm before buying.
Who it is best for
AWS data lake teams using the Glue Data Catalog and Glue ETL who want rule-based quality checks and anomaly detection without another vendor.
Typical use cases
- Checking completeness and uniqueness of S3 data lake tables registered in the Glue Data Catalog on a schedule
- Quarantining records that fail validation inside a Glue ETL job before they reach curated tables
- Detecting unusual drops in row counts with dynamic rules and anomaly detection
- Routing data quality failures to EventBridge for notifications or automated workflows
Who should look elsewhere
- Multi-cloud or multi-warehouse estates that want one observability tool with lineage: see Monte Carlo.
- Teams that run Spark outside AWS Glue and want the same rules in code: see Deequ.
Verdict
AWS Glue Data Quality is the natural first option for quality checks on an AWS Glue data lake: there is nothing to install, rules are portable DQDL, and costs follow DPU time. Its scope is the AWS Glue world, though, and it is rule-led, so wider estates or teams that want automatic monitoring with lineage usually pair it with, or choose, a dedicated observability platform.
Frequently asked questions
Is AWS Glue Data Quality a separate service?
No. It is a feature of AWS Glue, used from the Glue Data Catalog or inside Glue ETL jobs, and billed in Glue DPU-hours. See the AWS Glue profile.
How much does AWS Glue Data Quality cost?
There is no licence fee. Data Catalog recommendation and evaluation runs are billed per DPU-hour (at least 2 DPUs, 1-minute minimum), ETL checks add to the job's DPU time, and Data Catalog anomaly detection costs 1 DPU per statistic for the detection time. The AWS pricing page examples use USD 0.44 per DPU-hour.
What is the relationship between AWS Glue Data Quality and Deequ?
AWS states that Glue Data Quality is built on the open source Deequ framework from AWS Labs, and Deequ itself can run DQDL rules. That keeps rules in an open language you can also run on your own Spark.
Sources
- AWS Glue Data Quality (Developer Guide)
- Data Quality Definition Language (DQDL) reference
- AWS Glue pricing (Data Quality section)
- AWS Glue Data Quality product page
- Anomaly detection in AWS Glue Data Quality
Checked November 2026.
Listed in: Data Quality & Observability.
How we research reviews and what star ratings mean: our editorial method.