Great Expectations

Overview

Great Expectations is the best-known name in data testing, and 2026 split it in two, so read the ownership note before anything else.

GX Core, the part that matters now, is a Python library giving a programmatic interface for building and running validation workflows. It is Apache 2.0 and free, and the site says it will always remain free.

Checks are Expectations, verifiable assertions written as Python objects, for example expecting a column's values not to be null. They group into suites, a suite pairs with a batch to make a validation definition, and a checkpoint runs the lot and fires actions.

Data Docs are its signature feature and nothing else in this category matches them: GX renders human-readable documentation of your suites and results as a static site, so tests double as documentation.

Its check catalogue is the widest here, 73 modules covering completeness, uniqueness, ranges, aggregates and set membership, plus genuinely statistical expectations such as Kolmogorov-Smirnov, chi-square and KL divergence.

Execution is against SQL warehouses through published dialects, including BigQuery, Databricks, Fabric, SQL Server, PostgreSQL, Redshift and Snowflake, with anything else handled through SQLAlchemy, plus pandas and Spark.

Batch definitions partition an asset by day, month or year so you validate a slice rather than a whole table, though dataframes can only ever be taken whole.

Now the ownership. GX Cloud was acquired by FICO and stopped being publicly available on 1 June 2026, and GX Core passed to Fivetran as steward, whose GitHub organisation now holds the repository.

Be careful reading the vendor site: the GX Cloud marketing pages still advertise a managed service with no withdrawal notice, contradicting the company's own blog. The blog is the accurate source.

Release cadence is healthy, roughly fortnightly, with 1.23.1 shipping in September 2026 and recent work on Oracle SQL, Trino and ClickHouse test harnesses and agent skills.

Key features and capabilities

The same headings are used for every testing & validation entry, so two tools can be read side by side.

How checks are written

• Expectations are Python objects instantiated from the expectations module and added to a suite
• Suites persist in a data context, so they are reusable between sessions
• A validation definition pairs a suite with a batch definition, and a checkpoint runs it in production
• Authoring is Python-first; no YAML or SQL surface for expectations is documented in version 1

What it can check

• 73 expectation modules ship in the core library
• Completeness, uniqueness and set membership, including distinct values in, containing or equalling a set
• Aggregates and ranges covering minimum, maximum, mean, median, standard deviation, sum and quantiles
• Statistical expectations unique among these tools, including Kolmogorov-Smirnov, chi-square and KL divergence
• Cross-column pair checks; freshness and cross-table referential integrity are not published

Where checks run

• Published SQL dialects include BigQuery, Databricks, Fabric, SQL Server, PostgreSQL, Redshift and Snowflake
• Any other SQL dialect is handled through SQLAlchemy
• In-memory validation against pandas and Spark data sources
• Credentials come from environment variables or an uncommitted config file, referenced by variable
• Whether validation is pushed down and run as SQL in the warehouse is not stated explicitly in the docs

Results and alerting

• Data Docs render suites and results as human-readable documentation, unique to GX in this category
• Only a filesystem backend is documented for Data Docs; S3, Google Cloud Storage and Azure Blob are not published
• Actions fire from a checkpoint, with Slack notification and Data Docs update the two documented
• Email, Teams, Opsgenie and PagerDuty actions are not published for version 1
• Result verbosity is configurable, with checkpoints defaulting to a summary rather than complete output

Scale and performance

• Batch definitions partition an asset by day, month or year, or take the whole table
• Dataframes cannot be partitioned and are always retrieved whole
• Trino data sources do not support query data assets
• Version 1.20 added support for running a file data context on a read-only filesystem
• Sampling and an incremental validation mechanism are not published

Integrations

• An official Airflow provider for validating data directly from a DAG
• A published dbt tutorial covering Postgres, dbt, GX and Airflow together
• Spark through the Spark data source, plus Databricks SQL and Microsoft Fabric as dialects
• Dagster and a continuous integration recipe are not published in the docs
• No public REST API, since GX Core is a library

Where it runs and what it costs

• A library, installed with pip, on Python 3.10 to 3.13
• GX Cloud was the managed option and is no longer publicly available as of 1 June 2026
• The repository now sits in Fivetran's GitHub organisation following the stewardship transfer
• Regions for the former cloud product are not published
• No paid tier is currently on sale

Pricing

Open sourceFree, Apache 2.0

GX Core is free under Apache 2.0, and the site commits to it always remaining free. The commercial side has gone: GX Cloud was acquired by FICO and stopped being publicly available on 1 June 2026, so there is currently no managed GX product to buy. The pricing page still references a free developer option plus team and enterprise plans but publishes no figures, currency, units or trial length. Whether FICO will resell the cloud product, and at what price, is not published.

Vendor pricing page →

Demos and videos

• GX Core 1.0 demo Product demo · YouTube
• Unpacking data quality in feature engineering Webinar or talk · YouTube

About Fivetran

Great Expectations began as an open-source project with the company Superconductive behind it, which raised $21m to develop it. In 2026 the two halves separated: FICO acquired GX Cloud and withdrew it from public sale on 1 June 2026, while Fivetran announced on 13 May 2026 that it would become steward of the open-source community and the GX Core project, intending to fund maintenance, integrations and community work and to hire engineers with GX experience. Neither deal's value is published, and what became of the Superconductive entity is not published either. GX Core remains Apache 2.0 with roughly 404 contributors.

Founded 2017 · greatexpectations.io