Proactive Anomaly Detection

Find what you didn't know to look for

Traditional monitoring watches for known problems. Engagor AI scans across dimensions you're not watching — ISP x Device x Geo x Identity — finding correlations buried in the noise. The anomalies that matter most are often the ones you didn't know to look for.

Multi-dimensional pattern analysis. Cross-correlation detection. Proactive discovery.

ISP DEVICE GEO IDENTITY Gmail x iOS promo-eu DE x Mobile Tablet + SCAN 24/7
Critical
Warning
Info

Beyond single-metric thresholds

Most alerting systems watch individual metrics in isolation. "Bounce rate > 5% -> alert." But the most dangerous problems emerge from combinations — patterns that only appear when you cross-reference ISP with device, or geo with identity. That's where Engagor AI excels.

ISP Dimension

Performance by mailbox provider. Gmail vs Microsoft vs Yahoo patterns. Provider-specific throttling and filtering behaviors.

Device Dimension

Mobile vs desktop vs tablet engagement. Device-specific rendering issues. Platform distribution shifts.

Geographic Dimension

Country-level performance. Regional ISP variations. Timezone-correlated engagement patterns.

Identity Dimension

Per-sender identity health. Stream isolation. Brand and campaign performance comparison.

Cross-Dimensional Correlation

Where dimensions intersect, insights emerge

The most actionable findings come from combinations. Here's what cross-dimensional analysis reveals that single-metric monitoring would miss.

ISP x Device
Critical

Gmail + iOS engagement collapse

Open rates normal overall, but Gmail on iOS devices specifically dropped 34%. Indicates Apple Mail Privacy Protection filtering issue — not a sender problem.

Why this matters: Aggregate metrics hide device-specific problems. You'd chase the wrong root cause.
Geo x Identity
Warning

Germany underperforming on promo-eu

DE open rate 12.4% vs 21.6% average. Only affects promotional stream, not transactional. Suggests content or timing issue specific to German promotional campaigns.

Why this matters: Identity-level problems masked by overall geo performance. Actionable because it's specific.
ISP x Time
Warning

Microsoft deferrals spike at 2pm UTC

Deferral rate 3x higher between 14:00-16:00 UTC specifically at Microsoft. Indicates throttling during peak sending window.

Why this matters: Timing the problem enables timing the solution — shift volume away from peak.
Device x Geo
Info

France mobile engagement surge

FR mobile clicks up 28% while desktop flat. Shift in audience behavior or successful mobile-optimized campaign in French market.

Why this matters: Opportunity detection, not just problem detection. Double down on what's working.
Multi-Dimensional

The intersection is where problems hide

Global metrics look fine. But when you slice by ISP × Device × Geography × Identity, the anomalies emerge. That's where Engagor looks.

CORRELATE ISP / PROVIDER Gmail Yahoo Microsoft Telenet Orange Free.fr METRIC Open Rate Bounce DEVICE Mobile Desktop Tablet Webmail GEOGRAPHY US UK FR DE BE NL ES IT TIME COMPARISON Yesterday Baseline SENDER IDENTITY transactional.acme.com promo.acme.com newsletter.acme.com
DETECTED ANOMALY
-23%
OPEN RATE
Identity promo.acme.com
Device Mobile
Geography Germany
ISP Telenet

All other dimension combinations remain within normal baseline. Issue is isolated to this specific intersection.

HIGH CONFIDENCE
Detection Types

What Engagor AI looks for

Spike Detection

Sudden increases in bounce, deferral, or complaint rates. Day-over-day comparison catches single-day spikes that period averages would dilute.

Telenet.be 17% bounce (vs 0.9% baseline) — 18.3x spike

Decay Detection

Gradual engagement decline over time. 7-day vs 7-14 day comparison reveals trends that daily snapshots miss.

Open rate -25% over 14 days — gradual fatigue pattern

Distribution Shift

Changes in device mix, geo distribution, or ISP share. A 10-point mobile share drop might indicate rendering or deliverability issues.

Mobile share 45% -> 32% (-13 points) — investigate

Correlation Detection

Patterns that only emerge when crossing dimensions. Engagor AI finds relationships between seemingly unrelated metrics.

Complaint + correlates with volume + on Tuesdays

Threshold Breach

Absolute thresholds for critical metrics. Complaint rate >0.10% is dangerous regardless of baseline — immediate alert.

Complaint rate 0.12% — above critical threshold

Baseline Deviation

Learned baselines per identity, per ISP. What's normal for your promo stream differs from transactional — the system knows.

2x above YOUR baseline (not arbitrary threshold)
Engagor AI Analysis

Pattern detection that thinks

Beyond rule-based detection, Engagor AI analyzes your aggregated data looking for patterns that don't fit predefined rules. It asks: "What's unusual here that the thresholds wouldn't catch?"

Finds correlations across unrelated metrics
Identifies emerging trends before they hit thresholds
Provides natural language explanations
Generates confidence scores for each finding
Suggests root causes and recommended actions
Engagor AI 87% confidence

Emerging pattern detected

"Your Gmail bounce rate is normal overall, but when isolated to mobile devices in DACH region, there's a 2.3x elevation starting 3 days ago. This pattern preceded a major throttling event for another customer last month."

Recommendation:

Monitor closely for 48 hours. Consider reducing DACH Gmail volume preemptively if trend continues.

Proactive, not reactive

Traditional monitoring tells you when something broke. Engagor AI tells you when something is about to break — and often, why.

Reactive Monitoring
  • Alert when metric > threshold
  • Single-dimension only
  • False positives from noise
  • Manual root cause analysis
  • Damage already done
Proactive Detection
  • Alert when pattern emerges
  • Multi-dimensional analysis
  • Learned baselines filter noise
  • AI-suggested root causes
  • Prevent damage before it happens
Get Started

Ready to find what you're missing?

See how multi-dimensional anomaly detection can uncover patterns hiding in your email ecosystem.