ClickHouse vs MongoDB

Quick verdict

ClickHouse is an open source, column-oriented database for analytics, under the Apache License 2.0, self-hosted or run as ClickHouse Cloud. The latest stable release line is 26.9. Its storage, key and update model are covered in ClickHouse vs PostgreSQL; this page is about analysing document data.

MongoDB is a document database storing BSON documents in collections, queried with the MongoDB Query API and aggregation pipelines, with replica sets, sharding and multi-document transactions. The current manual release is 9.0. Community Server is under the SSPL v1.0, and MongoDB, Inc. runs the Atlas cloud service.

Side by side

Key differences

Analytics in MongoDB: the aggregation pipeline and its limits

The MongoDB manual describes an aggregation pipeline as one or more stages that process documents, each passing its output to the next. It handles filters, grouping, window functions and joins with $lookup, and $merge or $out can write results to a collection, which is a common way to maintain summary collections for dashboards.

// MongoDB (mongosh): revenue by country for shipped orders
db.orders.aggregate([
  { $match: { status: "shipped" } },
  { $group: { _id: "$customer.country", revenue: { $sum: "$total" }, orders: { $sum: 1 } } },
  { $sort: { revenue: -1 } },
  { $limit: 10 }
])

The limits are documented. A stage that needs more than 100 MB of memory either writes temporary files to disk or fails, depending on allowDiskUseByDefault and the per-command allowDiskUse option, and MongoDB 9.0 adds a per-operation memory limit (by default 1 GB or 20% of the memory available to the server process, whichever is greater). A pipeline is limited to 1000 stages. Heavy pipelines also run on the same replica set as your application, which is why Atlas offers analytics nodes: replica set members, available on dedicated M10 and larger clusters, that never become primary and are targeted with the nodeType:ANALYTICS read preference tag. In our view, that combination is enough for many operational dashboards and should be tried before adding a second database.

Documents in ClickHouse: the JSON type

ClickHouse's documentation states that the JSON data type was marked production ready in open source release 25.3 and is not recommended in production on earlier versions. It splits each document into subcolumns, one per path, up to max_dynamic_paths (1024 by default); further paths go into a shared structure that is less efficient to query. Paths without a type hint are read as Dynamic values and can be cast; paths you know in advance can be given a type, and unwanted paths skipped.

-- ClickHouse 26.9: orders as JSON, with type hints for paths you query often
CREATE TABLE orders_json (
    order_id String,
    doc      JSON(status String, total Decimal(12,2), customer.country String, SKIP internal)
) ENGINE = MergeTree
ORDER BY order_id;

SELECT doc.customer.country AS country, sum(doc.total) AS revenue, count() AS orders
FROM orders_json
WHERE doc.status = 'shipped'
GROUP BY country
ORDER BY revenue DESC
LIMIT 10;

-- a path without a type hint is Dynamic; cast it when reading
SELECT doc.channel::String AS channel, count() FROM orders_json GROUP BY channel;

Because each path is stored as its own compressed column, a query that touches three fields reads three columns rather than whole documents, which is the point of moving analytics to ClickHouse. The trade-off is that ClickHouse is not an operational document store: primary keys do not enforce uniqueness, updates are handled differently from MongoDB's in-place document updates, and documents arriving from MongoDB must be deduplicated by version, as described on our ClickHouse vs MySQL page for replicated rows.

Getting MongoDB data into ClickHouse

ClickHouse documents three routes, and they do different jobs:

• MongoDB ClickPipe (ClickHouse Cloud only). ClickHouse's documentation states that ingesting from MongoDB via ClickPipes "is in public beta"; the Cloud changelog dates the promotion to public beta to January 2026. It performs an initial load and then continuous change data capture using MongoDB's Change Streams, which rely on the oplog. Supported sources are MongoDB Atlas, self-hosted MongoDB and Amazon DocumentDB, with MongoDB 5.1.0 or later; the beta announcement added sharded clusters, PrivateLink, SSH tunnelling and X.509 authentication. Documents are replicated into ClickHouse's JSON type by default.
• MongoDB table engine and mongodb() table function. A read-only link to a remote collection, supported for MongoDB 7 and later. Simple WHERE, ORDER BY and LIMIT conditions are pushed down to MongoDB; nested documents and arrays arrive as JSON strings. Suitable for one-off loads and small lookups, not for keeping a copy in sync.
• Your own pipeline for self-hosted ClickHouse: read MongoDB change streams with a CDC tool, often through a message queue, and write into ClickHouse tables that handle updates and deletes, for example with ReplacingMergeTree and a version column.

Because the ClickPipe is still a beta, check its documentation for current limitations, plan for resyncs, and make sure the oplog retains enough history for the pipe to catch up after an interruption.

Which fits which workload

MongoDB is the right home for the documents themselves: an order, a user profile, a product with variable attributes, read and updated by key or index with transactions where needed. ClickHouse is designed for the questions asked across millions or billions of those documents and the events around them: revenue by day and country, funnel analysis on clickstreams, percentiles over logs, joins with data from other systems.

A useful test is where the pain is. If application latency suffers when reports run, isolate them first (Atlas analytics nodes, secondaries, or $merge into summary collections). If reports themselves are too slow or too hard to express as pipelines, or analysts want SQL and BI tools, that is the case for ClickHouse. For a wider view of the database-versus-warehouse decision, see data warehouse vs database.

Pricing and licensing

ClickHouse open source is free under the Apache License 2.0. ClickHouse Cloud bills per compute unit-hour and per TB-month of storage, varying by plan, cloud and region. ClickHouse's CDC billing page states that MongoDB ClickPipes are free during public beta and that billing starts when the connector reaches general availability; the same page lists, for CDC ClickPipes, USD 0.10 per GB for initial load or resync and USD 0.20 per GB for continuous replication, plus CDC compute of USD 0.20 per hour on Scale and Enterprise (USD 0.10 on Basic).

MongoDB Community Server is free under the SSPL. MongoDB Atlas, listed on the MongoDB pricing page in October 2026 in USD: Free tier with 512 MB of storage; Flex at 0.011 per hour, capped at 30 per month; Dedicated clusters from 0.08 per hour (M10). Analytics nodes are available only on dedicated (M10 and larger) clusters, so isolating reports in Atlas means a Dedicated cluster; use the Atlas pricing calculator for the extra nodes.

Pricing checked on the vendors' official pages on 7 October 2026. Prices change; confirm before buying.

Where each one leads

Limitations

When to choose each

When neither is right

• Your data is relational and you need joins and constraints in the operational store: see MongoDB vs PostgreSQL.
• You want a fully managed warehouse rather than an analytics database: see ClickHouse vs BigQuery and ClickHouse vs Snowflake.
• You need a cache in front of MongoDB rather than analytics: see Redis vs MongoDB.
• Your analytics are local or file-based: see DuckDB vs PostgreSQL.

Final recommendation

Frequently asked questions

Sources

• ClickHouse documentation: JSON data type
• ClickHouse documentation: MongoDB ClickPipe
• ClickHouse blog: MongoDB CDC connector for ClickPipes is now in Public Beta
• ClickHouse Cloud changelog 2026
• ClickHouse: ClickPipes for CDC billing
• ClickHouse documentation: MongoDB table engine
• ClickHouse GitHub releases
• MongoDB manual: Aggregation operations
• MongoDB manual: Limits and thresholds
• MongoDB Atlas: High availability and workload isolation (analytics nodes)
• MongoDB Atlas: Pre-defined replica set tags
• MongoDB Atlas pricing

Checked October 2026.

How we research comparisons: our editorial method.