EP 14: What is Observability?

Summary

This Data Science With Sam episode has Sam interview Ed Ferron of Exigent Solutions about Observability as the next operating layer beyond siloed monitoring. The discussion connects Application Performance Monitoring, Full Stack Observability, Business Transaction Observability, Observability Security Telemetry, AI-Enabled Observability, OpenTelemetry, Proactive Observability, and Real-Time Operational Analytics. Its core synthesis is that application telemetry becomes more valuable when it explains end-to-end customer experience, business impact, security risk, and infrastructure cost in real time.

Key Claims

• Observability is framed as a major jump from monitoring, not just another version of monitoring.
• Traditional logs, infrastructure metrics, database checks, queue metrics, network data, and cloud-service dashboards can stay too siloed to explain end-to-end application behavior.
• Full Stack Observability should let teams move from a business symptom to technical root cause across application, network, database, cloud, security, and other layers.
• Business Transaction Observability translates telemetry into stakeholder language such as order flow, ride-order latency, affected customer percentage, and affected locations.
• The source treats application performance as a revenue and customer-experience issue, including B2B cases where short delays can have large financial consequences.
• Observability Security Telemetry is becoming part of the observability stack because vulnerabilities, bad libraries, bad DLLs, and attacks can interrupt customer onboarding, login, and other business functions.
• AI-Enabled Observability is useful because developers and tools emit too much data for humans to manually correlate across systems.
• Machine-learning methods can help detect anomalous relationships among signals and attach context to alerts.
• Proactive Observability matters because customers often retry, restart, abandon a service, or complain on social media before filing a formal support ticket.
• OpenTelemetry is presented as an important standard for generating, ingesting, querying, modeling, and alerting on observability data.
• For data scientists, metrics, events, traces, logs, and spans can support Real-Time Operational Analytics without replacing enterprise reports or data warehouses.
• Observability can support scaling and cost decisions during demand spikes such as Black Friday, back-to-school periods, and electric-vehicle charging use cases.
• Ed argues that teams often already own strong observability tools but have not tied them to the business activities they need to understand.

Key Quotes

“next major jump in monitoring” - Ed’s framing of observability’s relationship to earlier monitoring.

“order a ride is slow” - business-language example for customer-visible telemetry.

“as much data as possible” - Ed’s description of the observability data-ingestion posture.

Connections

• Data Science With Sam, Sam (Data Science With Sam), Ed Ferron, and Exigent Solutions - show, host, guest, and organization.
• Observability, Full Stack Observability, Business Transaction Observability, and Application Performance Monitoring - core monitoring-to-observability frame.
• OpenTelemetry and Real-Time Operational Analytics - telemetry standard and data-science use case.
• AI-Enabled Observability, Proactive Observability, and Production ML Feedback Loops - signal-correlation, alerting, and operational feedback branch.
• Observability Security Telemetry, Cybersecurity Data Science, and Security Data Access Constraint - security telemetry and risk-management connection.
• MLOps, Machine Learning Engineering, and Data Engineering For Data Science - adjacent production-data and model-operations concepts.

Contradictions

• No direct contradiction found.
• The source extends EP 7: Data Science & MLOps by moving from model deployment feedback into whole-application telemetry, business transactions, and cost-aware operations.
• The source qualifies Cybersecurity Data Science by showing that security signals are not only separate threat datasets; they can also be part of business-facing observability when attacks or vulnerable components degrade customer workflows.