INTEGRATION · CLICKHOUSE

Real-time product analytics on ClickHouse.

Klaritics on ClickHouse is the fastest configuration we ship. ClickHouse's vectorized columnar engine and MergeTree architecture deliver sub-second p95 latency at high concurrency — ideal for real-time dashboards, customer-facing embedded analytics, and event volumes that would buckle other warehouses.

Deploy Klaritics Read the ClickHouse setup guide →

✅ GA · Fastest p50Latency (typical):<500ms p50, <2s p95Min version:ClickHouse 22.8+Compression:10–30× columnarDeployment:Self-hosted (yours or ours)

How Klaritics queries ClickHouse

Klaritics connects to your ClickHouse cluster (self-managed, ClickHouse Cloud, or Altinity Cloud) via a read-only user. Queries from the UI compile into ClickHouse SQL, taking advantage of features other warehouses don't have:

- **Vectorized execution** — operations on entire blocks of column values at once - **Sparse primary indexing** — instead of indexing every row, ClickHouse indexes ranges, making sub-second range scans practical - **Aggressive pruning** — partitions, primary keys, and skip indexes combine to read only the data the query touches - **Materialized views (`AggregatingMergeTree`)** — pre-aggregate retention and funnel data in real-time as events stream in - **Lightweight DELETE** — fast soft deletes for GDPR / right-to-be-forgotten workflows

For high-concurrency dashboards, Klaritics' query planner generates SQL that's tuned for ClickHouse's strengths — flat column access, push-down predicates, and pre-aggregated materialized views where appropriate.

Why teams pair Klaritics with ClickHouse

Sub-second analytics at any concurrency

For sub-second p95 analytical queries on fresh data with 50+ concurrent users, real-time OLAP databases like ClickHouse are the only architecture that physically supports the workload. Cloud DWHs hit latency, cost, and concurrency limits at this profile.

Storage economics that are hard to beat

ClickHouse's columnar compression typically achieves 10–30× compression on event data — meaning the same event archive that's 10TB on Postgres is often under 1TB on ClickHouse, with much faster scans.

Customer-facing analytics done right

If you're embedding product analytics inside a customer-facing dashboard (think: a SaaS giving its users their own analytics view), ClickHouse is purpose-built for this. Hundreds of concurrent dashboard queries returning in under a second is its happy path.

Where it shines (and where it does not)

Where it shines

Real-time and near-real-time product analytics (sub-minute freshness)
High-concurrency dashboards (100s+ of concurrent queries)
Customer-facing embedded analytics
Very large event volumes (10B+ rows) where storage economics matter
Use cases requiring deep retention (compression makes 5+ year retention affordable)

Where it does not

Frequent updates and deletes (ClickHouse's mutations are async and heavyweight). Best for append-mostly workloads.
Complex multi-table joins on huge tables (modern ClickHouse handles joins well, but a star-schema BI workload may still favor Snowflake).
Teams without operational experience running self-managed clusters — ClickHouse Cloud or a managed provider eases this. Klaritics works with both.
Very small data (<10GB) — a managed ClickHouse cluster has fixed minimum cost; Postgres might be more economical at this scale.

Getting started

Step 1

Step 1 — Create a Klaritics read-only user

CREATE USER klaritics_reader IDENTIFIED BY 'use-a-secrets-manager';

Step 2

Step 2 — Grant read access to your event database

GRANT SHOW, SELECT ON analytics.* TO klaritics_reader;

Step 3

Step 3 — (Recommended) Settings profile to limit query resources

CREATE SETTINGS PROFILE klaritics_profile SETTINGS
  max_memory_usage = 10000000000,  -- 10 GB per query
  max_execution_time = 60,         -- 60 seconds
  readonly = 1;
ALTER USER klaritics_reader SETTINGS PROFILE 'klaritics_profile';

Step 4

Step 4 — Connect from Klaritics admin UI

# Settings → Warehouses → Add ClickHouse
# host: clickhouse.internal
# port: 9440 (TLS) or 8443 (HTTPS)
# database: analytics
# user: klaritics_reader
# ssl: require

Sample query: 7-day signup-to-activation funnel

WITH step_1 AS (
  SELECT user_id, event_time AS step_1_time
  FROM analytics.events
  WHERE event_name = 'signup'
    AND event_time >= now() - INTERVAL 30 DAY
),
step_2 AS (
  SELECT s1.user_id,
         min(t.event_time) AS step_2_time
  FROM step_1 s1
  INNER JOIN analytics.events t
    ON t.user_id = s1.user_id
   AND t.event_name = 'activated'
   AND t.event_time BETWEEN s1.step_1_time
                        AND s1.step_1_time + INTERVAL 7 DAY
  GROUP BY s1.user_id
)
SELECT
  uniqExact(s1.user_id) AS signed_up,
  uniqExact(s2.user_id) AS activated,
  uniqExact(s2.user_id) / nullIf(uniqExact(s1.user_id), 0) AS conversion_rate
FROM step_1 s1
LEFT JOIN step_2 s2 ON s1.user_id = s2.user_id;

💡 for production funnels with a fixed step list, ClickHouse's `windowFunnel()` function is dramatically faster than the self-join pattern. Klaritics' planner uses `windowFunnel` automatically when the funnel definition allows.

Security and compliance

Read-only user with `SELECT` on the analytics database; nothing more.
TLS connections only (port 9440 native protocol or 8443 HTTPS).
Settings profiles to cap memory, execution time, and rows read per query — protecting your cluster from runaway dashboards.
Row-level security via ClickHouse's `CREATE ROW POLICY` works transparently with Klaritics — combine with Klaritics' team permissions.
Compatible with self-managed ClickHouse, ClickHouse Cloud, Altinity, Aiven, and DoubleCloud.