Precomputing fraud signals with ClickHouse materialized views at Moniepoint

Johnson Adekunle and Abraham Imohiosen, Engineering Managers at Moniepoint, explain how precomputed customer state with ClickHouse materialized views turns fraud checks into lookups, reducing p90 aggregate lookup latency by 93.1%.

Moniepoint needs to evaluate transactions quickly enough to act before fraudulent money leaves the system. Johnson Adekunle and Abraham Imohiosen, Engineering Managers at Moniepoint, explain how they moved from computing customer state for each transaction to precomputing it with ClickHouse materialized views.

Their architecture uses ClickHouse, a cache writer and a key-value store so the evaluator can look up customer state. With the same traffic profile and rule set, aggregate lookup latency fell by 85.3% at p50, 93.1% at p90 and 82.1% at p99.

Recorded at Open House London on 30 September 2026.