My Data Quality Tool Market Guide

What Is a Data Quality Tool?

A data quality tool is a solution to data reliability problems that implements features such as data tests, data profiling, data observability monitors, data quality workflows, data quality dashboards, data lineage, and incident management.

Can an AI code assistant build a data quality tool from the definition above? Not today.

I wrote this guide while curating a list of 30+ specialized data quality tools. Examples of data quality tools featured include data testing tools (Great Expectations, Soda, DQOps, Deequ, AWS Glue Data Quality, Google CloudDQ, DQX), data observability tools (Elementary, Monte Carlo, Metaplane, Bigeye, Anomalo, DQLabs, Lightup, Validio, Telmai, Acceldata, Unravel, Pantomath, IBM Databand, Coalesce Data Quality (formerly SYNQ), DataKitchen, Kensu, Masthead, FirstEigen, Qualdo, CluedIn, iceDQ), shift-left data quality tools (Datafold, Foundational, Gable, Recce, Entropy Data), and unified data trust tools with a data catalog (Sifflet, Decube, Ataccama, RightData, SelectZero, OpenMetadata, Collate, OpenDataDiscovery).

Data tools from other product categories that also implement data quality features will be covered separately. Notably, data governance tools (Collibra, Atlan…), MLOps/LLMOps tools (Evidently AI, Langfuse…), data warehouses (Snowflake, Databricks…), and data platforms (dbt Cloud, Qlik…).

Data Quality Tool Features

Data quality is traditionally thought of in terms of building on the understanding of your data and, consequently, building trust. To understand data, you need to look at your data. Yes, you cannot scan all your data, one table, one row, and one column value at a time. But you can efficiently look at a data sample, profile your data, build expectations, and test these expectations to grow your understanding, one test result at a time. Or you can rely on AI to do all this for you, can you?

What follows is an enumeration of the most popular data quality tool features, with examples from a few tools I consider implement each feature best.

Data Testing

Data tests and monitors are the most common features implemented by data quality tools. Tests and monitors compare data to expectations. While data tests require hardcoding expectations, data monitors compute expectations based on historical values to detect anomalies in recent data.

This section focuses on data testing tools and features to test SQL tables in data warehouses, in-memory DataFrames with Python, and unstructured data in object stores with LLMs. Data observability tools and features are covered later.

SQL Table Tests

There are a few open-source data testing tools developed by companies that also offer managed solutions. Notably, you may have heard of Great Expectations (GX Core, GX Cloud), Soda (Soda Core, Soda Cloud), and DQOps (DQOps Cloud).

It’s a bit tricky to draw the line between the open-source projects and the managed solutions. For example, GX Core provides a different developer experience than GX Cloud. While GX Core is fundamentally a Python library of 50+ data expectations that can be configured as JSON objects to be passed as parameters to a Python function, GX Cloud is a managed solution with a web UI that allows you to apply data expectations to data sources. Soda Core is a CLI and Python library compatible with the Soda Checks Language (SodaCL) to write data quality tests and data contracts in YAML based on 25+ built-in data metrics and execute them programmatically. But to access more advanced features, such as data observability monitors, you need to install the Soda Library, which requires a Soda Cloud account. I find it easier to read and write tests as YAML rather than JSON. Thus, I consider Soda to be more user-friendly for those looking for an open-source code-based data testing library.

A third option is DQOps, which implements 150+ built-in data quality checks as templated Jinja2 SQL queries that can be configured through YAML or a web UI. Note that to store data quality definitions and results in the cloud or create data quality dashboards, you need a DQOps Cloud account. A fourth option is DataKitchen's DataOps testing and observability open-source projects, with 40+ types of data quality tests with paid team and enterprise plans.

Each tool names similar tests differently, making it hard to compare test coverage. Most tools would implement at least built-in checks for not null, unique, column types, and support custom SQL tests. Custom SQL tests take the form of SQL queries that return failing rows or are used to flag a table or column.

You will find many other, more or less popular solutions by searching on GitHub. Previous and ongoing open-source data testing tools include MobyDQ (inactive) by the data team at Ubisoft, Datavines, and pointblank by the creators of RStudio.

Otherwise, data platforms have implemented their native data testing solutions, one after the other. Amazon open-sourced AWS Deequ and sells AWS Glue Data Quality, Databricks open-sourced DQX for Spark and announced Agentic Data Quality Monitoring (powered by Unity Catalog), Google open-sourced CloudDQ for BigQuery (and now offers it as part of Dataplex Universal Data Catalog), and Snowflake built DMFs (data metric functions) and may expand data quality features as part of Snowflake Horizon. Not to forget dbt Core and Cloud, which are a common starting place for data teams to add built-in dbt tests and contracts (more on data contracts later), dbt-utils tests, dbt-expectations tests, and custom SQL tests.

Several tools categorize tests by data quality dimensions. While not all tools agree on the list of data quality dimensions most will include at least accuracy, validity, consistency, uniqueness, completeness, and timeliness.

Data Contracts

Data contract tools like Entropy Data, Soda, Foundational, Gable, and Collate (managed OpenMetadata platform) define data contracts as a way to formalize expectations between data producers and data consumers.

I like to think about data contracts as a generalization of data tests (traditionally focused on data schemas and validation) to define groups of expectations and metadata for each data asset. This include: data schemas, data validation rules, data ownership, data documentation, SLAs, data policies...

You may have also heard about dbt contracts. While dbt test checks whether data meets a condition, a dbt contract prevents data from being materialized unless it meets the contract. To me this feels like bringing back good old data schemas (with a confusing name) which we abandoned when data professionals transitioned from databases to data warehouses that don’t require data schemas to be defined beforehand. Note that not all data contract tools will automatically prevent failing records from landing in the destination and quarantine it in intermediary tables for validation. This depends on how you integrate and run data contracts.

There is one main Open Data Contract Standard (ODCS). Entropy Data (previously known as Data Mesh Manager or Data Contract Manager) supports and contributes to ODCS and allows testing data contracts with their open-source Data Contract CLI to check data contracts during CI/CD or periodically. Note that Entropy Data is a solution focused on building data product marketplaces.

Soda’s data contracts is a Python library to write data contracts in YAML using a specific format, but can also build contracts from ODCS files. Each time the pipeline accepts or produces new data, you can programmatically trigger Soda data contracts or leverage Soda Cloud.

Collate supports writting data contracts in YAML that should follow their data contract JSON schema.

Gable’s data contracts are defined in YAML using a proprietary specification and stored in a separate git repository. Note that Gable integrates with databases and cloud storage (not data warehouses) and it's focused on validating data objects in memory.

Data Comparison Tests

This section covers data comparison tests, notably: data-diff tests, data comparison checks, and data reconciliation tests. Data diff tests compare all values across two datasets. Data comparison is more widely supported and most often is limited to checking that the number of rows or primary keys is the same between two datasets. Data reconciliation tests are used to validate data migration or data replication jobs.

A full-featured data diff tool reports table-level, column-level, row-level, and value-level data differences. Data diff tools include Datafold and Recce. One common pattern between these tools is to be able to compare data between production, staging, and development environments to validate SQL/ETL code changes based on data impact reports during CI.

Datafold’s Data Diff compares datasets fast, within or across databases and files. In the past, Datafold open-sourced the data-diff CLI with a dbt integration, now archived and forked by one of the original creators as reladiff. Now, Datafold focuses on automating and validating data warehouse and ETL migrations with AI agents that translate SQL and validate code changes based on data-diff reports. That, an cost optimization and data knowledge graphs to automate data engineering.

Recce is a dbt-specific tool to streamline the process of validating dbt code changes by seeing data impact reports in your PR, with full-featured data-diff views. It also provides a lineage diff graph.

Soda supports three types of reconciliation checks: metric reconciliation checks, record reconciliation checks, and schema reconciliation checks. Ataccama’s data reconciliation feature provides high-level checks of data consistency across multiple sources, mostly used for validating data migrations.

Bigeye supports join rules for validating data across tables and sources and deltas to validate successful data replication, migration, or model code changes. Telmai implements data diff for Parquet files.

DataComPy by Capital One is an open-source library to compare two DataFrames (or tables) such as Pandas, Spark, Polars, or Snowflake.

DataFrame Tests

DataFrame tests are a means to validate data in files or database tables by loading it into a DataFrame, or to validate data in transit in your data pipelines.

Deequ by Amazon is an open-source library built on top of Apache Spark to define unit tests for any data that you can fit into a Spark DataFrame, like CSV files, database tables, logs… You can write tests in Scala, Java, Python with PyDeequ or DQDL (Data Quality Definition Language). Amazon also offers AWS Glue Data Quality, a managed solution built on top of Deequ. Databricks open-sourced DQX, a Python data quality framework for validating PySpark DataFrames across all Databricks engines: Spark Core, Spark Structured Streaming, and Lakehouse Pipelines / DLT.

The once-popular Apache Griffin project is now archived. Be aware of what can happen when an open-source project is not backed by a company. While once promising (albeit not a DataFrame testing library), OpenDataDiscovery faced a similar challenge with 10 months in 2025 with no commits on their Github repo.

Other open-source alternatives to validate Spark DataFrames include the independent SparkDQ library and Spark Expectations by Nike. DataFrame-agnostic libraries include Cuallee (Snowpark, PySpark, BigQuery, Pandas, DuckDB, Polars, Daft) and Union Pandera (Pandas, Polars, Dask, Modin, Ibis, PySpark).

Data Code Tests

Data code tests leverage SDKs and CLI's to validate data objects in memory from your codebase. Source code analysis is the process of finding the dependencies of code functions on data assets.

Gable static code analysis recognizes code (Python objects, PySpark, TypeScript) that generates data payloads, capturing the structure and types as a re-usable data asset definition. Data asset definitions can then be used to create data contracts. Gable also supports registering assets for Protobuf messages, Avro records, JSON Schema files, and S3 files.

Foundational has integrations for popular ORMs (SQLAlchemy, Liquidbase, Entity Framework Core, TypeORM, Ruby Active Record) to extract lineage by analyzing your source code and then creating contracts.

IBM Databand Python SDK allows logging and monitoring custom metrics and Python/Java/Scala function invocations. The Elementary Python SDK allows you to programmatically report data quality metrics to Elementary Cloud from your Python data pipeline code.

While not a data quality tool per se, Pydantic is a popular data validation library that leverages Python type hints.

Unstructured Data Tests

Unstructured data tests verify that unstructured data conform to specific rules. Rules are created as text prompts and applied to data through an LLM.

Elementary’s unstructured data validations validate unstructured data in a data warehouse field using an LLM prompt. Elementary Data leverages warehouse-native AI functions: Snowflake Cortex AI LLM functions, Databricks AI Functions, and Vertex AI models for BigQuery. This functionality can be virtually leveraged by any data quality tool that supports custom SQL tests, as it’s advertised by Validio’s custom SQL validator with LLM function calls, Monte Carlo’s unstructured data monitoring, or Ataccama ONE’s integration with Snowflake Document AI.

Only Anomalo and Lightup, both specialized data monitoring tools, seem to have built specific UXs for validating unstructured data. Anomalo’s unstructured data monitoring leverages a library of prompts with a focus to flag documents that contain PII and redact them. Lightup provides AI Data Profiling for documents (type, size, summary, FAQ) and implements document-level and folder-level metrics for document stores (S3, Google Drive, Box, OneDrive). Lightup even open-sourced lightudq without much public success. I feel like we are pretty early here.

Data Profiling

Data profiling enables understanding the properties of data by computing key table and column-level metrics. Table-level profiling metrics include: schemas with column types, row count, column count, latest updated date… Column-level metrics include counts and % of null/not-null values, unique/duplicate values, distinct values (cardinality), max/min/avg/std of numeric values or of lenghts for strings.

Profiling results are used to create data tests and monitors, understand data models when writing aggregation logic, and help investigate a data quality issue. Most data quality and governance tools implement profiling. Here are some examples.

Monte Carlo data profiler profiles up to 20M records and samples data up to 10k rows to recommend monitors. On top of metrics, you can see histograms of row counts and common values with day-aggregated metrics. Atacama profiling provides graphs for frequency analysis, frequency groups, and quantiles. Non-numeric attributes also benefit from mask analysis, pattern analysis, and length statistics. Telmai implements interactive profiling with pattern detection, value distribution, and drill down.

You will also find several open-source data profiling projects on Github. ydata-profiling is the most popular data profiling library for Pandas and Spark DataFrames. Dataprof is a new and fast Rust library and CLI for profiling tabular data.

Data Observability

Data monitors are the core feature of data observability tools. Data monitors normally rely on data profiling results and anomaly detection algorithms to detect anomalies.

Data Monitors

Most data observability tools combine tests and monitors, while users can choose between manual and automatic thresholds. Monte Carlo, like Sifflet, lists data tests and monitors under one list.

Sifflet provides automatic monitors to monitor all tables within selected schemas or databases for: freshness, volume, and schema change. Otherwise, Sifflet provides over 20 advanced monitors to be configured for: table-level health, metrics, field profiling, format validation, and custom monitors.

Metaplane makes it easy to quickly add monitors at the database and table level through their UI. At the database and schema level, you'll see a full list of tables that you can add freshness, column count, row count, or custom SQL monitors to. At the table level, you'll be able to apply any of the monitor types that Metaplane supports.

Schema Monitors

Schema change monitors detect changes in data schemas, like the addition or removal of columns, or type changes.

Freshness Monitors

Freshness monitors detect changes in the frequency of updates to data tables, and normally only check table metadata, such as the last updated timestamp.

Volume Monitors

Volume monitors detect changes in the ratio of data updates, based on storing statistics about the number of row additions, updates, and removals.

Field Monitors

Field statistics monitors compute statistics on a field, like the count or % of duplicates, null values, wrong format values, or aggregations like the max, min, or avg. While tests include a hardcoded threshold, monitors track a metric and alert when there is a statistically significant change to that metric.

Custom Monitors

Custom monitors are normally written in SQL and compute business metrics that may aggregate multiple column values, and are used to alert of unusual activity, which can still be legit but should not happen unnoticed.

Window-Based Monitors

Validio provides window-based validators for data lakes, data warehouses, and data streams. Configuring a validator requires configuring a metric, a field, a filter, a window, segmentation, and a threshold. Validio supports tumbling and sliding windows. Validio backfills allow you to see incidents detected on the past data, which can be used to triage incidents to retrain the algorithms for future incident detection. Data quality scores are calculated as the ratio of windows without incidents to the total number of windows. Other tools that put forward window-based monitors include Telmai and Lightup.

Anomaly Detection

When creating anomaly detection tests, you first need to collect metrics for historical values. To configure anomaly detection tests, you need to specify at least a column with a timestamp, a frequency to compute buckets, and a sensitivity to fire alerts. Most often, seasonality and trends in your data are taken into account. When a tool says that it supports anomaly detection, it would do it for at least three metrics: freshness, volume (row counts), and schema changes. Supporting anomaly detection for custom metrics is more rare.

Pipeline Monitors

Pipeline monitors allow you to collect statistics (schedules, durations, costs) about pipeline jobs (or tasks), query logs, and database transactions. IBM Databand pipeline monitoring can create a lineage graph that spans tasks across dbt, Spark, Airflow, and custom code-based functions.

Data Quality Workflows

Data quality workflows allow you to schedule and orchestrate tests from your data pipeline code, data orchestrator, CI/CD pipeline, terminal, code editor, or AI agent.

Scheduling

All managed data quality tools to schedule tests periodically with standard frequencies or cron expressions.

Orchestration

Data quality orchestration is the process of triggering tests and monitors from other components of your data stack. This includes data orchestrator and data pipeline steps that trigger tests after loading or transforming data. This is useful when data is not produced or consumed at regular intervals, and you want to check for data quality more granularly.

Data testing tools like Soda and Great Expectations provide Python SDKs that allow you to create tasks to trigger test suites from data orchestration tools like Airflow, Prefect, or Dagster. You can also pull test results to semaphore pipelines.

DQX by Databricks applies checks on a dataframe and can quarantine invalid records to an intermediary table to ensure that “bad” data is never written to the output.

iceDQ uses the term data monitoring to gather services such as white-box monitoring (circuit breakers), data contracts, input data validation, and data reconciliation.

CI/CD

Datafold, Recce, Gable, and Foundational are part of the shift-left data quality movement and provide CI integrations.

For example, Datafold and Recce allow to compare dbt models during CI/CD and generate a data impact report as a PR comment with statistics about the models and columns that will be impacted.

MCP Server

MCP servers from data quality tools can be leveraged by coding agents to generate tests, debug issues, and refactor issues directly from the IDE. Beyond that, the MCP Servers from data quality and observability tools that are aware of data schemas and relationships can be leveraged from AI coding agents to generate code that works with your data.

AI Agents

Many data quality tools, notably data observability ones (Sifflet, Elementary…), announced AI agents to debug and resolve data incidents on behalf of the user. Some tools have even pivoted to only provide AI agents for all common data quality problems, such as Acceldata.

Elementary Ella provides specific AI agents for test recommendation, issue triage and resolution, governance (create documentation, enforce policies, tag data), performance, and data discovery.

Data Quality Dashboards

Data quality tools implement different kinds of data quality dashboards that include data quality KPIs, data observability reports, data quality scorecards, data table health reports, incident management reports, custom reports, and job performance reports.

Data observability reports provide statistics about the number of tests (with test coverage) and test results (failing/passing). Elementary provides a data observability dashboard with KPIs and evolution graphs for test results, table health, and test coverage.

Data quality score reports include the data quality score for each data quality dimension applied at the data source, table, and column levels. DQOps calculates data quality KPIs as a percentage of passed data quality checks out of all executed checks. Elementary uses the same formula and allows customization of the mapping of tests to dimensions and the weights for each dimension. Instead, Decube calculates data quality scores as the ratio between the number of failed and scanned rows.

Data table health reports gather all scores for a table and its columns in one view.

Incident management reports include metrics about the total number of incidents, status breakdowns, and time to resolution. Decube provides an incident dashboard with statistics about all incidents, incidents assigned, incident levels, data contracts breached, and average time to close incidents. Monte Carlo includes a data operations dashboard with the number of data incidents and time to resolution.

Custom data quality dashboards create views with filters to focus on critical data assets, data sources to share with data producers, or data models to share with data consumers. Metaplane allows the creation of custom dashboards for specific roles. DQOps allows the creation of fully customizable dashboards on Looker Studio.

Performance reports include statistics about ETL jobs and database queries. Monte Carlo performance dashboard allows for investigating high-cost queries, slow dbt jobs, and slow Airflow DAGs. Elementary dbt performance report is focused on execution times for dbt models and tests.

On top of being able to see and compare generic data quality metrics, most data quality tools allow you to navigate, filter, and sort test results in a table view.

Data Lineage

Data lineage is a popular feature of data quality tools to extract and visualize data dependencies. Data lineage helps data teams and AI Assistants understand data models, prevent data quality issues with data impact reports during CI, debug data quality issues with Root Cause Analysis (RCA), find unused data to deprecate, and plan data migrations.

Most data quality tools implement data lineage; when they don’t (e.g., Great Expectations, Soda), they integrate with data governance tools (notably Atlan and Alation) that provide data lineage and a full-featured data catalog.

There are two main approaches to how lineage is built: query log parsing and manifest-based lineage.

Query log parsing is the most common approach. Tools connect to your data warehouse’s query history (e.g., Snowflake’s QUERY_HISTORY, BigQuery’s INFORMATION_SCHEMA.JOBS) and parse the SQL statements to infer which tables were read and written. Tools that rely primarily on query log parsing include Monte Carlo, Metaplane, Bigeye, OpenMetadata and Sifflet.

Manifest-based lineage uses a compiled dependency graph generated by your transformation tool — most commonly the dbt manifest.json, which explicitly declares every model’s upstream and downstream dependencies. This approach is highly accurate for dbt models but is limited to what dbt knows about. Tools that are dbt-native and rely on the manifest include Elementary, Datafold, and Recce.

In practice, most commercial tools combine approaches: they parse query logs for broad coverage and enrich with manifest or runtime data when available.

Column-Level Lineage

Column-level lineage expands table-level lineage to show dependencies between columns within a database. Data quality tools with column-level lineage let you fold and unfold dependencies from a single model and column. Column-level lineage is most often built from parsing SQL statements from query logs on your data warehouse or the dbt manifest when available. All data quality tools that implement lineage provide column-level lineage, apart from Lightup, which implements only table-level lineage.

Bigeye’s column-level data lineage graph allows you to highlight the upstream and downstream dependencies of a column.

Validio runs daily jobs, for example, SQL queries, to collect current information about lineage relationships from the Information Schema. This means that a delay of a maximum of 24 hours can occur from when a new relationship is created until it is visible inside Validio.

Cross-Application Lineage

Cross-application lineage extends lineage from the data warehouse to multiple data storage locations, spanning data sources to BI dashboards and data activation.

Metaplane end-to-end data lineage shows dependencies for integrations across data integration (Fivetran, Airbyte), transformation (dbt), orchestration (Airflow), BI (Tableau, Looker…), and data activation (Hightouch, Census…).

Data Incident Management

Data incident management is the process of solving data quality issues, starting with triaging issues to create incidents, firing alerts when necessary, and resolving incidents.

Issue Triaging

Issue triaging is the process of classifying an issue and adding it to a workflow that will take it to resolution. Classifying an issue includes adding information such as the owner, priority, impact, the date the test started failing, correlation to other issues, and further context to help debug the issue, including reports about the evolution of the number of rows in the table and the last updated date. This way, owners can filter issues by data domain, sort by priority, and handle related issues in one go.

Elementary incidents start with a single event but can include multiple events grouped.

In Soda, when a check fails, you can create an incident in Soda Cloud to track your team’s investigation and resolution of a data quality issue.

Sifflet incident management also provides a timeline with a chronological view of the incident, including when the issue was first detected, when the incident was created, and any status changes or comments made by your team.

Alerts

Alerts are fired after some initial issue triaging. Alerts depend on notification configurations to notify owners across channels such as email, Slack, and Microsoft Teams. Some tools integrate with issue tracking software like JIRA (Metaplane) and ServiceNow (Soda).

Data Incident Resolution

When a test or monitor fails, there can be multiple causes and actions. Failing tests can be the cause of changes in the code, data, or infrastructure. Code changes break data tests, source data changes and schema changes break downstream data, and infrastructure (servers, ingestion jobs) may fail. Resolution may come from editing data, code, or infrastructure. Or issues may autoresolve on the next pipeline run or require manually rerunning a pipeline.

Monte Carlo provides Root Cause Analysis (RCA) insights and pull request correlation with automated findings to facilitate the discovery of the root cause of a particular data incident based on query logs, value correlations, and source code.

Bigeye’s BigAI provides incident descriptions, suggested resolutions, suggested preventions, and cron suggestions.

Some test failures are easier to interpret. The easier tests to analyze are those that flag a value in a row. Test results that depend on multiple rows or values (sum, average) or monitors that depend on dynamic thresholds are harder to interpret. You won’t immediately know which record needs to change for the test to pass. A test failure does not mean that there is something wrong with the data. In this case, some tools allow you to flag the data so that they stop getting false positives, and the anomaly detection algorithm can be fine-tuned.

Soda allows you to inspect rows that didn’t pass a test. Elementary allows you to provide false positive feedback and flag an anomaly as insignificant change, expected outlier, business anomaly, not an interesting table, or other.

Sifflet announced three AI agents (Sentinel, Sage, and Forge) to suggest monitors, debug issues, and suggest fixes. Sifflet Sentinel is an AI agent that suggests relevant data quality monitors to add.

Telmai implements data binning to categorize data into "good" and "bad" bins that need to be reviewed and circuit breakers.

Master Data Management

Master Data Management (MDM) is the process of data stewards solving data quality issues originating from source data (master data), and includes imputing missing data, removing duplicates, fixing inconsistencies, validating data changes, correcting data, enriching data (classification, extraction…), and merging records across sources.

Master Data Management (MDM) and Reference Data Management (RDM) at first sound more like legacy features from enterprise data management platforms like Informatica, but I am surprised that modern data quality tools haven’t implemented what I consider the most basic and important data quality feature, the ability to view and e dit your data. This feature today is more part of Data IDEs like DataGrip and DBeaver, and rising code IDEs for data teams (nao). I can’t believe that modern data teams using dbt and the like don’t ever curate master and reference data manually, apart from creating seed files.

Today, Ataccama is the data quality tool that best supports MDM, with complementary products: Ataccama ONE Data Quality & Catalog, Ataccama MDM, and Ataccama RDM. Ataccama ONE Data Tables provides a UX-friendly interface to navigate data tables with filters and data quality scores per column and flagged rows and columns. Ataccama MDM exposes master data through a web interface, with a rich feature set for data governance: browsing, searching, viewing, creating, and modifying data, and issue resolution support. Ataccama RDM manages reference data by supporting formal, defined processes and ensuring central authority over all reference data changes.

Data Quality Tool Integrations

Tell me who you integrate with, and I'll tell you who you are. I gathered integration support by tool in an Integrations Comparison Table with the help of Claude Code Agent Skills.

Monte Carlo is the most integrated data quality tool with over 60 integrations, followed by Collate, Anomalo, and Elementary. On the other side, you have platform-specific tools like DQX by Databricks, Google CloudDQ and AWS Glue Data Quality. Almost all tools integrate with the most popular data warehouses (Snowflake, BigQuery, Redshift, and Databricks) and most integrate with relational databases such as PostgreSQL and MySQL. Note that Gable is the only tool that doesn't integrate with data warehouses as it focuses on relational databases and data streams.

The second most popular integration category is BI tools (lead by Tableau and Power BI), which is mostly due to data quality tools that provide end-to-end data lineage, like Metaplane, Elementary, Monte Carlo, Bigeye, Sifflet, and Datafold. Specialized data testing tools like Soda and data monitoring tools like Anomalo promote integrations with data governance tools like Atlan and Alation to add data quality scores to catalog items and see data lineage.

dbt and Spark are the most popular data transformation integrations, followed by query engines such as AWS Athena, Trino, Hive SQL, and Presto. Data object and data lake integrations are unsurprinsigly lead by Amazon S3, Google Cloud Storage, Azure Data Lake Storage, and Databricks Delta Lake. Airflow outnumbers by far other data orchestration tools in term of integrations.

Data Quality Tool Pricing

Data quality tools range from free open-source projects to enterprise platforms costing hundreds of thousands of dollars per year. Most vendors don't publish prices publicly — but a handful do, and the AWS Marketplace surfaces list prices for enterprise tiers. Prices below are estimates based on public pricing pages and the AWS Marketplace, and may not reflect current rates. You can fill up the feedback form to suggest an update.

Where to start: DQOps, Soda Core, GX Core, DataKitchen, Elementary, and OpenMetadata all have free open-source versions you can self-host. Metaplane, SelectZero and Recce offer permanently free tiers for small teams. Kudos to Collate (managed OpenMetadata) for offering the most generous free plan (5 users, 500 tables) with more tables than in many other AWS Enterprise plans. Collate's team pricing offers the most competitive price per table at an easy to understand 1$/table/month rate.

On pricing models: I've noticed that most data quality tools update the pricing model (from user based, to monitor based, to table based, to consumption based) and cost several times per year (from public pricing, to private pricing, back to public pricing). Honestly, it feels like no tool has figured out a pricing that works, so they keep iterating from the easiest to implement model (user based) to the most complex one (consumption based). The good news is that you can use this to negotiate pricing, and ask for a custom price based on value. If you need historical data points, you can view older pricing by entering the pricing url in the Internet Wayback Machine. Good luck!

On enterprise pricing: The AWS Marketplace list prices for enterprise tiers tend to match the annual salary of a US-based Senior Data Engineer. These are list prices — vendors negotiate, and multi-year contracts typically get meaningful discounts.

Snowflake-native options: Some tools run entirely inside your Snowflake account without requiring credentials or data to leave the warehouse. Metaplane supports Snowflake native monitoring. Ataccama offers Data Quality Gates for Snowflake. Unravel has a Snowflake Native App for cost and performance optimization. Sifflet accepts Snowflake credits.

The Future of Data Quality Tools

Today, no specialized data quality tool implements all features across all integrations and use cases, so you have to make a choice. I hope this guide helped you better understand what to look for when choosing a data quality tool. You can check my list of 30+ specialized data quality tools where I add my opinion on which tool fits best who. Shortlist 1-3 tools and give them a try. Nothing have replaced, yet, building your own opinion.

As a bonus, ere are some data quality trends I see happening in 2026.

Data Quality and Governance Consolidatation

In 2026, I believe data observability and data governance tools will merge, with tools from each category already implementing features from the other. A few data observability tools have already implemented full-featured data catalogs, such as Sifflet, Decube, Elementary, and Bigeye. OpenMetadata/Collate and Ataccama ONE offer a data catalog from day one.

Data Quality Tools Rebrand as Context Tools

Data quality and governance tools already started to position themselves differently as a unified data trust platform (Decube), context layer for data and AI (Atlan), data and AI control plane (Elementary), data operations center (Pantomath), agentic data management platform (Acceldata)… Who will win?

For me, data quality main goal has always being to provide a solution to better understand data while data governenace helps act on it. Together data quality and governance build trust. Orignially for data people. Now context is what AI agents need to understand your data. Same story, different audience.

Data Quality Tools Get Acquired

Coalesce acquired SYNQ in March 2026 to launch Coalesce Data Quality, after they acquired CastorDoc for their data catalog in 2025. Will dbt follow and acquire Elementary? In 2025, Datadog acquired Metaplane, but these are still two separate products. In 2025, Monte Carlo raised $135M, Gable $20M, Sifflet $18M, Collate $10M, Datafold $4M, and Recce $4M. In 2026, Validio raised $30M and Decube $3M.

Snowflake is a bit short on data quality solutions, they acquired Select Star in 2025 to expand on Snowflake Horizon for data governance, but they only have Snowflake DMFs to build data tests and monitors. Google has Google Dataplex Data Quality (and Google CloudDQ), AWS has AWS Glue Data Quality (and Deequ) and Databricks open-sourced DQX and released Agentic Data Quality Management. Will Snowflake build or buy a data quality tool in 2026?

Data Quality Tools Pivot to Focus on ROI

AI agents closes the loop between understanding data and acting on it (vs doing nothing with it). Data quality is not anymore something that needs to be done necessarily before hand, but can be accomplished in paralel as you capitalize on your data. No need anymore to spend countless hours cleaning data that will never be queried. That poses the question whether data quality remains a separate category, or becomes furhter embedded on whatever data operatiosn layer we build for humans and AI.

For example, Datafold gave up on data quality, and they are now selling their data-diff technology for ROI-focused use cases: automating data stack migrations and optimizing data warehouse costs.

Data Quality Tools Add AI Observability

Today, Monte Carlo is the only data tool that implements AI observability features. Notably LLM observability, LLM monitoring and LLM evaluation. Monte Carlo went far enough to put AI Agent observability first on their homepage to the detriment of data observability. This is where the budgets are in 2026, anything AI related. Other's will likely follow. Eventually I see Data Quality, MLOps and LLMops tools merging as a combination of feature expansions, mergers and acquisitions.

What's next? Next, I am distilling the data governance category, followed by related categories such as MLOps/LLMOps, and data orchestration. I will then cover more established (and challenged by AI) product categories for data movement, transformation, BI and storage. Data platforms for the end.