Loading…
OBSERVABILITY · CLICKHOUSE · ANALYTICS
ClickHouse Behavioral Analytics Agent
Humans, bots, and agents on your site — visible, separated, explained.
THE CHALLENGE
Traffic isn't just "users" any more. Bots crawl. Other AI agents browse, click, and act. Treating the three populations as one stream gives you funnel numbers that hide the truth — and product decisions made on hidden truth.
THE APPROACH
An event pipeline on ClickHouse for scale, a classifier that separates the three populations (humans, bots, agents), and an agent layer on top that answers "where do they engage" and "where do they stop" in plain language — with the SQL it ran shown for review.
STACK / HOW IT WORKS
How it works.
ClickHouse
Columnar warehouse; sub-second over billions of events.
Client event SDK
Consistent schema across every page.
Population classifier
Splits humans / bots / agents by signal mix.
LLM SQL (fixed schema)
Parameterised SQL, explained; no ad-hoc DDL.
Materialized views
Cheap high-volume queries.
OUTCOMES
What it delivered.
Billions
Events queryable sub-second
3
Populations cleanly separated
100%
Queries shown for review
ARCHITECTURE
How it's built.
WORK WITH US
Want something like this built?
Tell us what you're trying to ship. A real engineer replies — no pitch.