Block · intelligent
Anomaly Time Series
Area chart that auto-detects anomalies via z-score and surfaces the top deviation as a plain-language narrative.
State
Size
Mode
Monthly recurring revenue
4,758,780
1 anomaly flagged at 2σ
- Revenue
- Anomalies
Revenue dropped to 3,950,000 on Feb 5, 2026, 2.6σ below a 4,500,741.717 baseline — driven by Stripe outage.
Acme · billing data · updated hourly
Demo data is illustrative. Replace with your own typed data prop.
How it computes
Z-score outlier detection: flag points whose standardized deviation exceeds a threshold.
zᵢ = (xᵢ − μ) / σ; flag when |zᵢ| ≥ threshold (default 2σ)Assumptions
- The series is approximately Gaussian.
- Anomalies are independent rather than clustered.
- The threshold is symmetric across both tails.
Honest about
- Threshold is an explicit knob — 1.5σ (loose), 2σ (standard), 3σ (strict) — so the sensitivity is your decision, not a hidden default.
- If every sample is identical, σ = 0 ⇒ all z = 0 ⇒ no false anomalies.
- No multiple-testing correction: with enough points some |z| ≥ 2 are expected by chance.