Alternative
Snowflake Horizon Catalog Alternative: Data Quality Monitoring Pricing and Lineage Limits
Snowflake Horizon Catalog is the governance layer built into Snowflake itself: discovery, tagging, classification, masking, the Trust Center, data quality monitoring through data metric functions, and column-level lineage over Snowflake objects. There is no separate license to buy, which makes it the sensible first stop for a Snowflake-only estate. The costs that surprise teams are the Enterprise Edition requirement, the serverless credits a scheduled data metric function burns, the one-year lineage retention window, and the fact that the documented lineage object list stops at the edge of Snowflake. Datatrail is read-only, connects in minutes, and follows the same columns out into dbt, the orchestrator and the BI layer where Horizon does not go.
Side by side
Datatrail vs Snowflake Horizon Catalog
| Capability | Datatrail | Snowflake Horizon Catalog |
|---|---|---|
| Column-level lineage inside Snowflake | ||
| Lineage into dbt, Airflow and BI dashboards | ||
| Lineage history kept longer than one year | One year | |
| Works without Enterprise Edition | Enterprise+ | |
| Impact analysis before you ship a change | ||
| Monitoring priced per seat, not per compute-hour | Serverless credits | |
| Covers warehouses other than Snowflake | ||
| Read-only, never moves your data |
Comparison reflects general product positioning and is provided in good faith. Verify current capabilities with each vendor.
See it live
Lineage and impact, self-serve
Read-only connection. Datatrail never moves or mutates your data.
What Snowflake Horizon Catalog actually is
Horizon Catalog is not a product you buy separately. It is the name Snowflake gives to the governance and discovery features already inside the platform, and the list is long: the Trust Center for misconfiguration monitoring, automatic sensitive data classification, object tagging, column masking and row access policies, the Internal Marketplace, semantic views, Cortex-generated descriptions, AI guardrails, the Iceberg REST Catalog, the Model Registry, and two capabilities most teams evaluate it for, data quality monitoring and column-level lineage.
Because it ships with the platform, the honest starting position is that a Snowflake-only team should use it. You are already paying for Snowflake. Turning on tagging, classification and the Trust Center costs nothing extra beyond serverless usage, and no third-party tool is going to enforce a masking policy better than the warehouse that owns the data.
The evaluation question is narrower than "Horizon or a vendor". It is: does your data actually stay inside Snowflake, and can you live with a one-year window?
What Horizon data quality monitoring costs, with the real numbers
Snowflake does not publish a price for Horizon Catalog because there is not one. The cost arrives in two places, and both are documented.
First, the edition gate. Snowflake's documentation states plainly that "Data Quality Monitoring requires Enterprise Edition", and the built-in lineage view is labeled an Enterprise Edition feature as well. If you are on Standard, the two features people most want from Horizon are the two you cannot use, and moving to Enterprise raises the price of every credit you burn, not just the governance ones. On AWS US East (Northern Virginia), On Demand list pricing is $2.00 per credit on Standard and $3.00 on Enterprise.
Second, the serverless meter. Data metric functions run on Snowflake-managed compute. The Snowflake Service Consumption Table effective 18 August 2026 sets the base serverless rate at one Platform Credit per Compute-Hour, where a Compute-Hour is defined as an XS virtual warehouse running for an hour, then multiplies it per feature. Data Quality Monitoring carries a multiplier of 2 for Snowflake-managed compute and 1 for Cloud Services. One footnote is worth reading twice: the ROW_COUNT data metric function uses a multiplier of 0.9, not 2, so the cheapest check you can schedule costs less than half of what every other check costs.
| Serverless feature | Compute multiplier | Enterprise, US East, per compute-hour |
|---|---|---|
| Data Quality Monitoring (most DMFs) | 2 | $6.00 |
| Data Quality Monitoring, ROW_COUNT DMF | 0.9 | $2.70 |
| Sensitive Data Classification | 0.9 | $2.70 |
| Trust Center | 1 | $3.00 |
| Serverless Alerts | 0.9 | $2.70 |
| Search Optimization Service | 2 | $6.00 |
Multipliers from the Snowflake Service Consumption Table, effective 18 August 2026. Dollar figures apply the On Demand Enterprise credit price of $3.00 for AWS US East (Northern Virginia) and exclude Cloud Services. Capacity contracts discount the credit price; the multipliers do not change. Verify against your own contract before budgeting.
Two details make the meter behave better than the headline suggests. Billing happens only when a scheduled DMF runs, so calling a metric ad hoc in a SELECT is free, and Snowflake caps you at 50,000 total DMF associations per account, which is a real ceiling if you were planning to blanket a large estate. The default schedule is once an hour, so the arithmetic that matters is checks multiplied by frequency multiplied by table size, not a per-seat number you can forecast on a spreadsheet at the start of the year.
Where Horizon lineage stops, in Snowflake's own words
Snowflake's built-in lineage is better than most people expect. It is genuinely column-level, it walks upstream and downstream one hop at a time, and it covers tables, dynamic tables, external tables, Iceberg tables, views, materialized views, semantic views, stages and ML objects such as datasets, feature views and models. It picks up lineage from COPY INTO, CREATE TABLE AS SELECT, CREATE TABLE CLONE, CREATE VIEW, CREATE MATERIALIZED VIEW, CREATE SEMANTIC VIEW, INSERT SELECT, MERGE and UPDATE.
The limits are documented rather than hidden, and they are the reason teams end up buying something on top:
- One year of history. Both column lineage and object lineage are retained for one year. An audit that asks where a number came from in 2023 is not a question the platform can answer.
- Objects that are not tracked at all: objects in a shared database, objects in the shared SNOWFLAKE database, and objects in the INFORMATION_SCHEMA. Temporary tables are not shown. Deleted tables are not shown, though renamed ones are.
- Movement that is not tracked. If a table is used only to filter or join and its data does not land in the downstream object, that relationship does not appear. Separate disjointed queries are not stitched together, and data movement inside a stored procedure needs the GET_LINEAGE function rather than the graph.
- The object list is a list of Snowflake objects. Read it again and notice what is absent: there is no dbt model, no Airflow task, no Fivetran connector, no Tableau workbook, no Power BI report, no Looker explore. Snowflake markets Horizon lineage as end to end, and inside the platform it is. The documented catalog of things it can draw is Snowflake objects and stages.
That last point is the whole decision. If the question you ask most often is "which dashboard breaks if I rename this column", the answer lives in a system Horizon does not model. Datatrail parses query history and the dbt graph into column-level lineage and then runs impact analysis across the exposures and dashboards downstream, before the change merges. We have written up the wider Snowflake picture in our guide to the Snowflake data catalog and in Snowflake lineage.
Honest guidance: when Horizon is the right answer
Use Horizon Catalog and nothing else if your estate is genuinely Snowflake-shaped: ingestion lands in Snowflake, transformation happens in Snowflake with dynamic tables or views rather than an external orchestrator, the people asking lineage questions are Snowflake users, and you are already on Enterprise. In that world the native tooling is free of an extra license, enforced at the engine, and impossible to drift out of sync.
Look at something else when any of these are true. Your transformation layer is dbt or Airflow and you need to see the job, not just the resulting table. Your consumers are in Power BI, Tableau or Looker and the question is always which report breaks. You keep lineage for compliance and one year is short. You run a second warehouse alongside Snowflake, in which case Horizon by definition sees half the estate. Or you are on Standard Edition and the upgrade to Enterprise for two features costs more than the tool would.
A fair way to frame it: Horizon is governance of Snowflake. What most data teams need on top is lineage across the stack. For the full field, see our comparison of data lineage tools and data quality tools, and the Atlan data catalog comparison if a full catalog is what you are really shopping for.
Questions people ask
Snowflake Horizon Catalog and Datatrail, answered
Is Snowflake Horizon Catalog free?
There is no separate license or SKU for Horizon Catalog, so in that sense it is included with Snowflake. The cost shows up in two other places. Data quality monitoring and the built-in lineage view both require Enterprise Edition or higher, which raises your credit price across the board, and scheduled data metric functions consume serverless credits at a multiplier of 2 on top of the base rate of one credit per compute-hour.
How much do Snowflake data metric functions cost?
Snowflake bills data metric functions as a serverless feature at one Platform Credit per compute-hour multiplied by the Data Quality Monitoring multiplier of 2, plus Cloud Services at a multiplier of 1. On the On Demand Enterprise price of $3.00 per credit in AWS US East, that works out to roughly $6.00 per compute-hour of checking. The ROW_COUNT metric is the exception at a multiplier of 0.9, about $2.70. You are billed only when a scheduled DMF runs, not when you call one in a SELECT.
Does Snowflake Horizon do column-level lineage?
Yes. Snowflake's built-in lineage traces relationships between columns in source and target objects and lets you expand upstream or downstream one level at a time. It requires Enterprise Edition, covers tables, views, materialized and dynamic tables, Iceberg and external tables, semantic views, stages and ML objects, and retains both column and object lineage for one year. Dashboards, dbt models and orchestrator tasks are not in the documented object list.
How long does Snowflake keep lineage history?
One year. Snowflake documents a one-year retention window for both column lineage and object lineage. If you need lineage evidence for an audit that reaches further back than twelve months, or you want to show how a metric definition changed over several years, you need to capture and store that history outside the platform.
What does Snowflake Horizon lineage not track?
Snowflake documents the gaps: objects in a shared database, objects in the shared SNOWFLAKE database, objects in the INFORMATION_SCHEMA, temporary tables and deleted tables are not shown. Relationships where data is only filtered or joined without moving into the downstream object are not captured, separate disjointed queries are not stitched together, and data movement inside a stored procedure requires the GET_LINEAGE function instead.
Do I still need a data lineage tool if I use Snowflake Horizon?
It depends on where your pipeline lives. If everything from ingestion to consumption happens inside Snowflake, Horizon covers it. If dbt or Airflow does the transformation, or people read the numbers in Power BI, Tableau or Looker, the lineage that answers "what breaks if I change this" runs through systems Horizon does not model, and you need a tool that reads across the whole stack.
What is the difference between Snowflake Horizon Catalog and Snowflake Open Catalog?
Horizon Catalog is the governance and discovery layer over the objects in your Snowflake account. Open Catalog is Snowflake's managed Apache Polaris service, an Iceberg REST catalog that lets external engines such as Spark, Flink and Trino read and write Iceberg tables. They solve different problems: one governs, the other brokers table metadata for open engines. Open Catalog is billed separately as a serverless feature at 0.5 credits per million requests.
Does Snowflake Horizon monitor freshness and schema changes?
Snowflake ships system data metric functions that cover common checks including freshness and row counts, and you can write custom ones in SQL. The default schedule is hourly, and you are capped at 50,000 DMF associations per account. What it does not do is rank an alert by what breaks downstream, because that requires lineage across the dbt and BI layers rather than lineage inside Snowflake.
Other comparisons
See it on your own warehouse
Connect read-only, transparent pricing, see your lineage in minutes. Datatrail never moves or mutates your data. Decide for yourself.