We compared 11 data catalogs by who carries the running effort: the platform vendor, the catalog vendor, or your own team. Platform-native catalogs, such as Unity Catalog and Google Knowledge Catalog (formerly Dataplex Universal Catalog), run inside a platform you already pay for. Atlan, an AI Context Platform, is vendor-hosted and spans platforms. DataHub Core and OpenMetadata ask your own team to run them.
Pick two or three from the list below, then run them through the shortlist tool with your own stack.
A few notes before the table:
- We order by who carries the running effort, from the vendor to your team, and alphabetically inside each group, so no product gets a position for being a sponsor. No vendor publishes a headcount line between “small” and “enterprise”, so that split is an editorial frame.
- Every entry is described on the same seven criteria, split across two tables so they stay readable on a phone.
- Every fact comes from the vendor’s own documentation, checked November 17, 2026. Where we found no documented fact, the cell says “see the vendor’s current docs” rather than guessing, and no release status is asserted anywhere.
| Field | Content |
|---|---|
| Query this answers | best data catalog for small teams vs enterprises |
| Entries compared | 11 data catalogs in three groups |
| Ordering rule | Who carries the running effort (platform vendor, catalog vendor, your team), then alphabetical inside each group |
| Current as of | November 17, 2026 |
| Status in vendor docs | Not asserted; see each vendor’s current docs |
| Who this is for | Data leaders choosing a first catalog, or replacing one that outgrew the team running it |
What does “small team” versus “enterprise” mean for a data catalog?
No vendor sets a headcount line between a small team and an enterprise, so the useful split comes from the vendors’ own sizing words. A data catalog here means the product that holds metadata, such as names, owners, definitions and lineage, so people can find data and govern it.
Two vendors size their offer in plain terms. DataHub says self-hosting Core gets you started. It points to Cloud once governance and security needs “outpace what a small team can maintain,” in its own Cloud versus Core comparison. Collate sizes its plans by users and data assets, and its Free Tier is labeled “For Small Data Teams” on its pricing page.
So “small” means few people to install, upgrade and keep a catalog running. “Enterprise” means more than one domain and platform, and a governance load that grows faster than headcount.
Is the category line itself the open question? Data catalog vs context layer explains what AI agents need beyond a catalog. Context layer vs data catalog vs semantic layer sets the three side by side.
Pick wrong and the cost lands on your team’s week: install, upgrades and troubleshooting for a catalog nobody has time to run, or a second platform the catalog cannot read. Read every product below through two questions first: who fixes it when an ingestion job fails, and who keeps owners and descriptions current after the launch?
Which data catalog fits which team? The comparison tables
Both tables describe all 11 entries on the same seven criteria, in the order of the rule: platform vendor first, catalog vendor second, your team last.
Table A: who runs it and how it starts
Table A covers who runs each catalog and how it starts.
| Catalog | People to run it | Setup route | Open source or managed | Native or cross-platform |
|---|---|---|---|---|
| Group 1: built into a platform you already run | ||||
| Databricks Unity Catalog | Databricks, inside your workspace | On by default for workspaces created after November 8, 2023 | Managed in Databricks; open-source implementation under Apache-2.0 | Native to Databricks |
| Google Knowledge Catalog | Google Cloud | A Google Cloud service; see the vendor’s current docs | Managed by Google Cloud | Native to Google Cloud; imports metadata from other platforms |
| Microsoft OneLake catalog | Microsoft, inside Fabric | Part of Microsoft Fabric | Managed by Microsoft | Native to Fabric; governs Azure and third-party sources through shortcuts, mirroring and connectors |
| Snowflake Horizon Catalog | Snowflake | Part of Snowflake; see the vendor’s current docs | Managed by Snowflake | Native to Snowflake; covers data inside or outside Snowflake |
| Group 2: vendor-hosted and cross-platform | ||||
| Alation | Alation Cloud Service, or customer-managed | See the vendor’s current docs | Cloud Service or Customer Managed deployments | Named by Microsoft as a Fabric integration |
| Atlan | Atlan, in a single-tenant SaaS tenant | Tenant on AWS, Azure or GCP; Self-Deployed Runtime add-on for running Atlan apps in your infrastructure | Managed, single-tenant SaaS | Cross-platform; named by Microsoft as a Fabric integration |
| Collate | Collate | Free Tier sign-up; paid plans start with a demo request | Managed, built on OpenMetadata | All plans include unlimited data source connectors |
| Collibra | See the vendor’s current docs | See the vendor’s current docs | See the vendor’s current docs | Named by Microsoft as a Fabric integration |
| DataHub Cloud | DataHub engineers | Guided implementation and onboarding | Managed | See the vendor’s current docs |
| Group 3: self-hosted open source | ||||
| DataHub Core | Your team, including uptime and troubleshooting | Manual install; Helm charts on Kubernetes for production; Docker quickstart for a local trial | Open source, Apache-2.0 | See the vendor’s current docs |
| OpenMetadata | Your team, across database, search and ingestion | Three production components with documented minimums | Open source; Apache-2.0 at the repository root, Collate Community License 1.0 on its ingestion and UI folders | See the vendor’s current docs |
Table B: price, reach and governance
Table B covers what each vendor documents about price, reach and governance.
| Catalog | How pricing is stated | Connector reach | Governance depth as documented |
|---|---|---|---|
| Group 1: built into a platform you already run | |||
| Databricks Unity Catalog | See the vendor’s current docs | Its metadata can be imported by Google’s Knowledge Catalog | Access control on queries and model calls, lineage tracking and audit logging, per Databricks docs |
| Google Knowledge Catalog | Pay-as-you-go, with free monthly usage on standard processing | Imports Unity Catalog, Snowflake Horizon, dbt, MySQL, Oracle, PostgreSQL and SQL Server | See the vendor’s current docs |
| Microsoft OneLake catalog | Included in Fabric capacity, not separate meters | Fabric, Azure and third-party sources | Microsoft’s data governance solution, per its November 12, 2026 post |
| Snowflake Horizon Catalog | See the vendor’s current docs | Iceberg ecosystem, including AWS Glue and Azure OneLake; lineage across BI tools and OpenLineage feeds | Classification, masking and row-access policies (Enterprise Edition or higher), role-based access control, Trust Center, access history |
| Group 2: vendor-hosted and cross-platform | |||
| Alation | Pricing page opens a conversation form | See the vendor’s current docs | See the vendor’s current docs |
| Atlan | Subscription by edition, cataloged assets and seats, plus Flexi Credits; set on the order form | Connectors reference lists extraction, lineage and enrichment per connector | SSO and SCIM, personas, purposes that apply policies and masking to tagged assets, event and access logs |
| Collate | Published plans by users and data assets | Unlimited data source connectors on every plan | SSO from Premium; audit logs on Enterprise |
| Collibra | Product page offers a demo request | See the vendor’s current docs | See the vendor’s current docs |
| DataHub Cloud | Contact sales; scoped to your use case | See the vendor’s current docs | Role-based and attribute-based access control; SOC 2 compliant, per DataHub |
| Group 3: self-hosted open source | |||
| DataHub Core | Free (open source) | See the vendor’s current docs | Roles plus platform and metadata policies, per DataHub’s docs; “basic access controls”, per DataHub’s Cloud comparison |
| OpenMetadata | Open source, free to download | See the vendor’s current docs | See the vendor’s current docs |
Each profile below expands one row with its sources. A cell that says “see the vendor’s current docs” means we could not confirm the fact on the vendor’s own pages on November 17, 2026. It does not mean the capability is missing, so ask the vendor before you score it.
How did we order these data catalogs?
We order by who carries the running effort, from the vendor to your team, and alphabetically inside each group, so no product gets a position for being a sponsor. Three groups follow from that rule.
- Group 1, built into a platform you already run. The data platform vendor runs the catalog. Databricks calls Unity Catalog the governance layer “built into Databricks,” and Microsoft says governance in Fabric is “included in Fabric capacity.”
- Group 2, vendor-hosted and cross-platform. The catalog vendor runs it. Atlan documents a single-tenant SaaS tenant, DataHub says Cloud is “fully managed by DataHub engineers,” and Collate runs every plan, up to private bring-your-own-cloud.
- Group 3, self-hosted open source. Your team runs it. DataHub tells Core users “your team owns uptime and troubleshooting,” and OpenMetadata documents three production components to size.
Atlan sits second in its group by the same rule, after Alation and before Collate.
The groups describe who runs the software, not who does all the work. A managed catalog takes install, upgrades and uptime off your team. Connecting sources, rotating credentials, assigning owners and keeping descriptions current stay with your team on every option.
The order says nothing about cost or quality. For dollars, the open source vs commercial total cost of ownership comparison scores both sides. To test whether a product is a real context layer or a relabeled catalog, or whether to build or buy metadata tooling at all, use those criteria next to this order.
Which data catalog is best for a small team?
For a small team, the right catalog is the one your people can run. Read the “People to run it” and “Setup route” columns first. Each group asks for a different share of your week.
Self-hosted open source puts install, upgrades and uptime on your team. According to DataHub’s quickstart documentation (2026), the tested local setup is 2 CPUs, 8 GB of RAM, 2 GB of swap and 13 GB of disk. That sizing is for a Docker trial on one machine. For production, DataHub’s Kubernetes guide lists four main components: GMS, two optional consumers and the frontend. They run on four external dependencies: Kafka, a database such as MySQL or Postgres, an Elasticsearch search index and a graph index.
OpenMetadata’s minimum requirements page (2026) documents three production components. The database needs 4 vCPUs and 16 GiB of memory. Search needs at least one master and two worker nodes. Airflow ingestion needs 4 vCPUs and 16 GiB.
Someone on your team owns each of those. What OpenMetadata is used for covers the project’s jobs, and DataHub pricing sets out what open source and Cloud cost to run.
A managed service moves install, upgrades and uptime to the vendor. Collate’s Free Tier is labeled “For Small Data Teams” and includes 5 users and 500 data assets on multi-tenant SaaS.
If your platform already ships a catalog
If you already run a data platform, its catalog may already be on.
- Unity Catalog is automatically enabled for all Databricks workspaces created after November 8, 2023, according to Databricks documentation (2026).
- According to Google Cloud’s Knowledge Catalog pricing (Google Cloud, 2026), pricing is pay-as-you-go, with free monthly usage on standard processing.
- According to Microsoft’s November 12, 2026 governance post (Microsoft, 2026), governance in Fabric is included in Fabric capacity rather than billed through separate meters.
If the goal is fewer data questions landing on your team, a data catalog for self-service analytics covers the five jobs a catalog does for everyone else.
If agents are on the roadmap, extending an existing catalog for AI readiness shows what to add later rather than buy now. Starting on open source and moving later is a path DataHub itself describes, from Core to Cloud. Migrating from DataHub to Atlan covers one route to a different managed catalog.
No vendor names a team size at which one group beats another. Count the people each candidate needs, then pick the one your team can still run after its busiest quarter.
Which data catalog is best for an enterprise?
For an enterprise, the right catalog is the one that reaches every platform and domain you govern, so read “Connector reach” and “Governance depth as documented” first. Several native catalogs now read beyond their own platform.
- According to Google Cloud’s metadata import connectors (Google Cloud, 2026), Knowledge Catalog imports from Databricks Unity Catalog, Snowflake Horizon, dbt Core and Cloud, MySQL, Oracle, PostgreSQL and SQL Server. Its Unity Catalog and Snowflake Horizon connectors support instances hosted on Google Cloud, Azure or AWS.
- According to Snowflake’s Horizon Catalog documentation (Snowflake, 2026), Horizon’s governance controls include sensitive data classification, masking and row-access policies, role-based access control, Trust Center and access history.
- According to Microsoft’s governance post (Microsoft, 2026), the OneLake catalog can govern Fabric, Azure and third-party data through shortcuts, mirroring and supported connectors. The post adds that “Fabric also integrates with other 3rd party providers such as Atlan, Alation, Collibra and Informatica.”
What vendor-hosted catalogs document for this lane
Vendor-hosted catalogs package for this lane in their own terms. Collate’s Enterprise plan is labeled “For Complex Use Cases & Enterprise-Wide Initiatives” and adds audit logs, dedicated support and private bring-your-own-cloud deployment. DataHub’s own comparison lists role-based and attribute-based access control and SOC 2 compliance for Cloud.
According to Atlan’s architecture reference (Atlan, 2026), Atlan runs as single-tenant SaaS, with dedicated compute and storage for each tenant. Collate lists single-tenant or hybrid SaaS from its Premium plan, and Alation documents a Customer Managed option. Comparisons such as Alation vs Informatica and Amazon Quick Suite vs Atlan for data catalog show how these trade-offs play out between named products.
Tooling is only part of the enterprise question. Scaling data governance across business domains, embedded data governance that lives in workflows and a written data governance operating model decide whether people use any catalog you buy. Keeping the native catalogs and adding one layer that reads them is another route. A data catalog that sits on top of existing catalogs explains that pattern.
Docs alone do not settle how a catalog behaves at your asset count, so test it in a pilot. Enterprise fit still comes down to documented reach: which platforms each candidate reads today, in its own docs, not on a roadmap slide.
Which data catalogs are built into a platform you already run?
The data platform vendor runs a platform-native catalog, so install, upgrades and uptime come with the platform you already run.
Each one covers its own platform first, and some now import metadata from others. Four entries follow, alphabetical.
Databricks Unity Catalog
Databricks runs Unity Catalog for you, inside the Databricks workspace. Databricks documentation (2026) calls it the “unified governance layer for data and AI built into Databricks” and says it is on by default for all workspaces created after November 8, 2023. The same page says it enforces access control when you query a table or call a model, tracks lineage and logs activity for auditing.
Unity Catalog is also published as an open-source implementation under Apache-2.0 (GitHub, 2026). For how its price is stated, see the vendor’s current docs. Its metadata also travels in one documented direction: Google’s Knowledge Catalog can import it, including instances hosted on Azure or AWS.
Before you treat it as your whole catalog, check which Unity Catalog business glossary and domains features your workspace has. Teams still on Hive can read Unity Catalog vs Hive Metastore, and teams that also run Microsoft tools can compare Unity Catalog and Microsoft Purview.
Google Knowledge Catalog (formerly Dataplex Universal Catalog)
Google runs Knowledge Catalog as a Google Cloud service. According to Google Cloud’s Knowledge Catalog overview (Google Cloud, 2026), as of April 10, 2026, Dataplex Universal Catalog is now called Knowledge Catalog. The API, client library, CLI and IAM names remain unchanged.
Google states that Knowledge Catalog “pricing is based on pay-as-you-go usage,” metered on processing and metadata storage. Free monthly usage applies to standard processing only, not to the Premium Processing SKU. Its metadata import connectors read Databricks Unity Catalog, Snowflake Horizon, dbt Core and Cloud, MySQL, Oracle, PostgreSQL and SQL Server.
Google Knowledge Catalog: rename, pricing and limits covers the rename in detail. If BigQuery is the center of your estate, read which layer is authoritative for BigQuery; Unity Catalog vs Dataplex compares the two native scopes.
Microsoft OneLake catalog
Microsoft runs the OneLake catalog inside Microsoft Fabric. “Microsoft Fabric is now our data governance solution,” Microsoft wrote on November 12, 2026. Governance in Fabric is “included in Fabric capacity rather than billed through separate meters,” per the same post.
The same post says the OneLake catalog can govern data across Fabric, Azure and third-party sources “without copying or moving it,” through shortcuts, mirroring and supported connectors. It names Atlan, Alation, Collibra and Informatica as third-party providers Fabric integrates with, and it sets November 12, 2027 as the end of support for Purview Data Governance. Per the same post, Microsoft Purview itself stays focused on data security and compliance.
What the OneLake catalog does covers its discover, govern and secure experiences, and Microsoft Fabric governance separates what is built in from what you add. Teams on Purview Data Governance can check Purview Data Governance coverage and follow a five-step plan to migrate from Purview Data Governance. For a Fabric-first shortlist, see the best data catalogs for Microsoft Fabric.
Snowflake Horizon Catalog
Snowflake runs Horizon Catalog as part of Snowflake. It is “the agentic catalog for all your data, whether it is inside or outside of Snowflake,” in Snowflake’s own words.
It reaches the Iceberg ecosystem “including AWS Glue and Azure OneLake, through catalog-linked databases,” according to Snowflake’s Horizon Catalog documentation (2026). It also traces column-level lineage across Snowflake, external databases, BI tools and OpenLineage feeds. The same docs name sensitive data classification, masking and row-access policies, role-based access control, Trust Center and access history. Snowflake’s masking and row access policy pages mark both as Enterprise Edition features, so check your edition first. For how its price is stated, see the vendor’s current docs.
What Snowflake Horizon Context is and how Snowflake Horizon Context and the Atlan context layer work together go deeper. For open table formats, Apache Polaris vs Unity Catalog vs Glue vs Snowflake Open Catalog compares the Iceberg catalogs.
Which vendor-hosted data catalogs work across platforms?
In this group the catalog vendor runs the software and carries upgrades and uptime, and each product connects to more than one platform. Five entries follow, alphabetical: Alation, Atlan, Collate, Collibra and DataHub Cloud.
Alation
Alation documents two ways to run it: Alation Cloud Service, and Customer Managed deployments, which its docs call Classic Alation. A notice in Alation’s documentation says that from August 18, 2026, Cloud Service docs moved to a new site.
Alation sits in this group because Cloud Service is vendor-hosted; its Customer Managed option moves the running effort to your team. Alation’s pricing page opens a “Start a Conversation” form. Microsoft’s November 12, 2026 post names Alation among the third-party providers Fabric integrates with. For setup effort and full connector reach, see the vendor’s current docs.
Alation vs Informatica compares two enterprise catalogs by team fit, and what AIOS is explains how Alation’s newer product relates to its catalog.
Atlan
Atlan, an AI Context Platform, runs as a cloud-first, single-tenant SaaS platform. According to Atlan’s architecture reference (Atlan, 2026), each tenant has its own dedicated compute and storage, and Atlan supports AWS, Azure and GCP as hosting platforms.
Atlan is priced as “a platform subscription set by edition, cataloged assets, and seats, plus Flexi Credits for the work Atlan AI does,” according to how Atlan pricing works (Atlan, 2026). Cataloged assets are the main driver of platform price. Admins and Members are paid seats, and Guests are free.
Everyday cataloging, browsing, searching and viewing lineage “never uses Flexi Credits,” and your order form sets the prices. A Self-Deployed Runtime add-on runs Atlan apps in your own infrastructure, and Atlan’s pricing page opens a talk-to-sales form.
One customer’s own account of scale comes from Mastercard’s Chief Data Officer:
"Atlan's metadata lakehouse is configurable, intuitive, and able to scale to hundreds of millions of assets."
Andrew Reiskind, Chief Data Officer, Mastercard
That is one customer’s assessment, not a benchmark, so test your own asset count in a pilot.
Atlan’s connectors and capabilities reference (Atlan, 2026) lists each connector’s extraction, lineage and enrichment capabilities. On how that context reaches AI agents, see the MCP-connected data catalog, the agentic data catalog, context engineering and how to implement an enterprise context layer for AI.
Collate
Collate runs a managed service built on OpenMetadata, and its pricing page lists three plans by users and data assets. “All plans include unlimited data source connectors,” the same page states.
The Free Tier, “For Small Data Teams,” includes 5 users and 500 data assets on multi-tenant SaaS with basic authentication. Premium, “For Teams & Small Business,” includes 25 users, 5,000 data assets, SSO integration and single-tenant or hybrid SaaS.
Enterprise, “For Complex Use Cases & Enterprise-Wide Initiatives,” includes 50 users and 10,000 data assets, each with an unlimited option, plus audit logs, dedicated support and private bring-your-own-cloud deployment. The open-source project underneath is covered in what OpenMetadata is used for.
Collibra
Collibra appears in Microsoft’s November 12, 2026 post among the third-party providers Fabric integrates with. Collibra’s Data Catalog product page directs buyers to request a demo.
For deployment options, how its price is stated and setup effort, see the vendor’s current docs. If you run Databricks next to it, Unity Catalog vs Collibra compares what each one governs.
DataHub Cloud
DataHub runs DataHub Cloud. DataHub’s Cloud versus Core comparison (2026) says Cloud is “fully managed by DataHub engineers” with SLA-backed availability. It also lists “guided implementation and onboarding” for Cloud.
Cloud pricing is “scoped to your use case(s) and data environment,” and buyers “contact sales for pricing,” per the same page. DataHub’s own comparison lists role-based and attribute-based access control and SOC 2 compliance for Cloud.
DataHub pricing prices open source and Cloud side by side, and DataHub alternatives compares nine options.
Which self-hosted open-source data catalogs exist?
A self-hosted open-source catalog is free to download, and your team runs it: install, upgrades, uptime and troubleshooting.
It gives you the most control and asks the most of your time. Two entries follow, alphabetical.
DataHub Core
Your team runs DataHub Core. DataHub’s Cloud versus Core comparison lists its cost as “Free (open source).” Its operating load reads “Manual install, configuration, and upgrades. Your team owns uptime and troubleshooting,” in the same comparison.
The DataHub repository (GitHub, 2026) is licensed Apache-2.0. DataHub’s quickstart is tested at 2 CPUs, 8 GB of RAM, 2 GB of swap and 13 GB of disk, which sizes a local Docker trial rather than production. For production, DataHub’s Kubernetes guide adds Kafka, a database, an Elasticsearch search index and a graph index as dependencies. DataHub’s own comparison lists basic access controls for Core, and its authorization docs describe roles plus platform and metadata policies.
Does running it no longer fit your team? Signs it is time to move off DataHub lists ten limitations to check. Then choose between DataHub Cloud and another managed catalog.
OpenMetadata
Your team runs OpenMetadata. Its README says the project is released under the Apache License 2.0. The root license file of its GitHub repository (GitHub, 2026) is Apache-2.0, while the ingestion and UI folders carry the Collate Community License Agreement 1.0. Read both before you build on it. Its minimum requirements page documents three production components, each with its own sizing:
- Database: 4 vCPUs, 16 GiB of memory and at least 100 GB of storage (150 to 200 GB recommended), on MySQL 8.0.42+ or PostgreSQL 15+.
- Search: at least one master and two worker nodes, each with 2 vCPUs, 8 GiB of memory and 100 GiB of storage, on Elasticsearch 9.x or OpenSearch 3.x.
- Ingestion: Apache Airflow at 4 vCPUs, 16 GiB of memory and 100 GiB of storage for DAGs and logs.
Collate offers a managed service built on the same project, with plans sized by users and data assets.
When do you need a catalog beyond the one your platform ships?
Consider a catalog beyond your platform’s own when your data sits on more than one platform. The test is whether the catalog you have gives every team the coverage, definitions and lineage it needs.
A native catalog covers its own platform first. The native catalogs now read past their borders, which narrows the gap without closing the question.
- Google’s Knowledge Catalog imports Unity Catalog and Snowflake Horizon metadata, including instances on Azure or AWS.
- Snowflake documents Horizon reaching AWS Glue and Azure OneLake through catalog-linked databases.
- Microsoft says the OneLake catalog governs Fabric, Azure and third-party sources, and names Atlan, Alation, Collibra and Informatica as third-party providers Fabric integrates with.
Importing another platform’s metadata is one step; giving every team and agent the same view of owners, definitions and lineage is another. Check where each candidate’s lineage stops at a platform edge, because that is where cross-platform questions start. Data lineage for AI traces AI answers back to source, and MCP for data lineage shows how agents query governed lineage.
The same choice shows up on every cloud. AWS Glue Data Catalog vs a cross-cloud governance layer makes it for AWS, and data marketplace vs data catalog separates publishing data products from finding data.
Native and cross-platform catalogs can run side by side. When the two disagree on an owner or a definition, which one will your people and agents treat as the answer?
How do these data catalogs state their prices?
Only some of these 11 catalogs publish a price, so the table records how each one states its price, not what it costs. Six patterns cover all of them.
| Pattern | Entries | What the vendor says |
|---|---|---|
| Free open source | DataHub Core, OpenMetadata | DataHub lists Core as “Free (open source)” under Apache-2.0; OpenMetadata’s repository root is Apache-2.0 |
| Pay-as-you-go | Google Knowledge Catalog | Pay-as-you-go, with free monthly usage on standard processing |
| Included in platform capacity | Microsoft OneLake catalog | Included in Fabric capacity rather than separate meters |
| Published plan sizes | Collate | Free Tier, Premium and Enterprise, sized by users and data assets |
| Order form or sales conversation | Alation, Atlan, Collibra, DataHub Cloud | Alation: a “Start a Conversation” form. Atlan: edition, cataloged assets and seats plus Flexi Credits, set on the order form. Collibra: a demo request. DataHub Cloud: “Contact sales for pricing.” |
| Platform pricing pages | Databricks Unity Catalog, Snowflake Horizon Catalog | See the vendor’s current docs |
A self-hosted catalog costs nothing to download. Its running cost is the people and servers behind its database, search and ingestion components, and no pricing page lists those for your estate.
For dollars rather than pricing models, the open source vs commercial total cost of ownership comparison scores ten options, and DataHub pricing prices one project both ways.
How do you test your shortlist before you buy?
Run two or three entries through the catalog shortlist tool with your own stack, then take the same questions into a written evaluation. The tool asks four questions that no table can answer for your stack, and it does not rank vendors:
- Connector coverage. Which of your systems does each candidate read today?
- Where lineage stops. At which platform edge does each candidate’s lineage end?
- Who maintains the metadata. Who keeps descriptions and owners current, and what makes them do it?
- What happens if they leave. Does the metadata survive the people who wrote it?
Put the answers in writing so every vendor answers in the same format. The data catalog RFP template turns them into vendor questions with weighted criteria. For governance-heavy buys, the 22 data governance platform tests add depth. If AI agents are part of the plan, evaluating an AI readiness platform covers what to test for them.
Copy the data catalog RFP template: 26 vendor questions and a 0 to 5 scoring sheet. Get the RFP template
What mistake should you avoid when choosing a catalog by company size?
The mistake to avoid is treating team size as the only input and skipping what sits outside your platform. That is our judgment from the vendors’ documented facts, not a claim any vendor makes.
Team size moves. DataHub frames its own Core-to-Cloud switch around the moment governance needs “outpace what a small team can maintain,” and Collate’s plans step from 500 data assets to 5,000 to 10,000. Platforms move too, and Microsoft’s own post names third-party providers Fabric integrates with.
Picture a five-person data team that picks a catalog for one warehouse, then inherits a second platform after an acquisition. The catalog was right for the team and wrong for the estate. Why data governance implementations fail traces how gaps like that surface, and whether your data catalog keeps up with what AI agents need adds the agent side.
Which platform will your team add in the next two years, and will the catalog you pick today read it?
FAQs about data catalogs for small teams and enterprises
Each answer below draws on the vendor documentation cited in the profiles.
1. What is the best data catalog for a small team?
The best data catalog for a small team is the one your people can run. If you already use Databricks, Google Cloud, Fabric or Snowflake, its native catalog runs inside the platform. Collate’s Free Tier, labeled “For Small Data Teams,” moves the running effort to the vendor. Self-hosted DataHub Core or OpenMetadata gives you control, and your team owns uptime.
2. What is the best data catalog for an enterprise?
The best data catalog for an enterprise is the one that reaches every platform and domain you govern. Compare documented connector reach and governance controls first. Knowledge Catalog imports Unity Catalog and Snowflake Horizon metadata, Snowflake lists masking and row-access policies, and Microsoft names Atlan, Alation, Collibra and Informatica as Fabric integrations. Need tenant isolation? Atlan documents single-tenant SaaS with dedicated compute and storage per tenant. Collate lists single-tenant or hybrid SaaS from Premium, and Alation documents Customer Managed deployments.
3. Is an open-source data catalog enough for a small team?
An open-source data catalog is enough when someone on the team can install, upgrade and keep it running. DataHub lists Core as “Free (open source)” and says your team owns uptime and troubleshooting. DataHub’s own comparison lists basic access controls for Core, and role-based and attribute-based access control for Cloud. Its Core docs also describe roles plus platform and metadata policies. Decide which controls you need first.
4. How many people does it take to run a self-hosted data catalog?
None of the vendors here states a headcount. Their docs state components instead. OpenMetadata documents a database, a search cluster with at least one master and two worker nodes, and an Airflow ingestion service, each with minimum sizing. DataHub’s Kubernetes guide runs its services on Kafka, a database, an Elasticsearch search index and a graph index. Someone on your team owns each piece.
5. Do I need a separate catalog if I already use Databricks, Snowflake or Google Cloud?
Consider a separate catalog when your data sits on more than one platform and your current catalog does not cover all of it. Unity Catalog is on by default for Databricks workspaces created after November 8, 2023. Knowledge Catalog imports Unity Catalog and Snowflake Horizon metadata, and Snowflake documents Horizon reaching AWS Glue and Azure OneLake. If no single catalog you run reads every platform you govern, test where each candidate’s lineage stops before you add one.
6. How do data catalogs state their prices?
The 11 catalogs here state prices in six ways. DataHub Core and OpenMetadata are free open source. Google bills Knowledge Catalog pay-as-you-go with free monthly usage on standard processing. Microsoft includes OneLake catalog governance in Fabric capacity. Collate publishes plans by users and assets. Alation, Atlan, Collibra and DataHub Cloud price through a sales conversation, and Atlan’s order form sets the price. For Unity Catalog and Snowflake Horizon Catalog, see the platforms’ own pricing pages.
7. Can I start with an open-source catalog and move to a managed one?
Yes. DataHub describes self-hosted Core as the way to get started, and Cloud as the option once governance and security needs outgrow what a small team can maintain. Collate offers a managed service built on OpenMetadata, with plans from 5 users up to an unlimited option. Before any move, plan how glossary terms, owners and lineage carry over.
8. Is Purview Data Governance still an option?
Purview Data Governance has a set end date. On November 12, 2026, Microsoft announced that Fabric is its data governance solution and that support for Purview Data Governance will end November 12, 2027. Microsoft Purview stays focused on data security and compliance, and governance in Fabric is included in Fabric capacity. Teams on Purview Data Governance should decide where their glossary, owners and policies go before that date.
9. What should I test before I buy a data catalog?
Test four things with your own stack: which of your systems each candidate connects to, where its lineage stops, who maintains the metadata and why, and what happens to that metadata if they leave. Then put the same questions into a written RFP with weighted criteria and a scoring sheet, so every vendor answers in the same format.
10. What is a data catalog?
A data catalog is the product that holds metadata, such as names, owners, definitions and lineage, so people can find data and govern it. The 11 entries here differ less in that job than in who runs them: the platform vendor, the catalog vendor or your own team.
11. Do I need a data catalog, or is it overkill for a small team?
You need one when people cannot find data or agree on what it means without asking the data team. If your data sits on one platform that ships a catalog, start there before you buy anything. A separate catalog earns its place when a second platform, domain or team needs the same definitions and lineage.
12. What is the difference between a data catalog and a context layer?
A data catalog helps people find and govern data. A context layer turns business knowledge, such as data definitions, working procedures and policies, into context AI agents can use. Atlan, an AI Context Platform, works as a context layer for AI across the platforms in this list.
Sources
- Knowledge Catalog overview, Google Cloud
- Knowledge Catalog pricing, Google Cloud
- Metadata import connectors, Google Cloud
- Snowflake Horizon Catalog, Snowflake documentation
- A next step in the evolution of the Microsoft Data Governance strategy, Microsoft Security Community Blog
- What is Unity Catalog?, Databricks documentation (docs.databricks.com)
- unitycatalog/unitycatalog repository, GitHub
- DataHub Cloud vs DataHub Core, DataHub (datahub.com)
- Quickstart guide, DataHub documentation (docs.datahub.com)
- datahub-project/datahub repository, GitHub
- Minimum requirements, OpenMetadata documentation (docs.open-metadata.org)
- open-metadata/OpenMetadata repository, GitHub
- Pricing, Collate (getcollate.io)
- Alation documentation and pricing, Alation (docs.alation.com, alation.com)
- Data Catalog product page, Collibra (collibra.com)
- Atlan architecture, Atlan documentation
- How Atlan pricing works, Atlan documentation
- Connectors and capabilities, Atlan documentation
- Pricing, Atlan
- Deploying with Kubernetes, DataHub documentation (docs.datahub.com)
- Policies Guide, DataHub documentation (docs.datahub.com)
- Understanding Dynamic Data Masking, Snowflake documentation
- Understanding row access policies, Snowflake documentation
- Access control, Atlan documentation