← All case studies
Case study · Financial risk
Tail-risk monitoring when correlations break
A multi-strategy asset manager rebuilt VaR and stress workflows on high-frequency, fat-tailed market data.
- Industry
- Asset management
- Profile
- Multi-strategy fund · $8B AUM
- Timeline
- Pilot in 38 days · BigQuery + internal OMS
+35%
VaR accuracy
90m
earlier alerts
99.9%
platform uptime
Summary
Risk committees were reacting to breaches after the fact. Predicta surfaced regime shifts and tail scenarios 1–2 hours earlier, with API-friendly outputs for the desk’s existing tooling.
The challenge
Overnight VaR runs missed intraday volatility clusters. Stress libraries were rebuilt manually when macro regimes shifted. Compliance wanted evidence trails — not another chart wall.
What we deployed
- 01Streamed BigQuery tick aggregates and internal position snapshots into Predicta’s regime-aware models.
- 02Calibrated fat-tail monitors with fund-specific liquidity constraints — not generic industry templates.
- 03Exposed threshold breaches through REST hooks into the existing OMS alert channel.
- 04Packaged stress narratives for the weekly risk committee in plain language.
BigQueryInternal OMSPredicta risk signalsTeams + email playbooks
Outcomes
- VaR back-test accuracy improved 35% versus the legacy batch workflow.
- Median alert lead time improved 90 minutes on high-severity tail events.
- Platform uptime held at 99.9% through two volatility spikes without manual failover.