New Relic

What is New Relic?

New Relic is a commercial, all-in-one observability platform. Where the open-source world assembles separate tools — Prometheus for metrics, Grafana for dashboards, Loki for logs, Jaeger for traces — New Relic bundles all of it into a single hosted product: application performance monitoring (APM), infrastructure metrics, distributed tracing, log management, real-user monitoring, and alerting, unified under one query language (NRQL) and one UI. The pitch is simple: instrument your app with their agent and get full-stack visibility without building or operating the pipeline yourself.

The real tradeoff: buy vs build

New Relic sits on one side of a decision every engineering org faces. Build the observability stack from open-source components (Prometheus, Grafana, Loki, Jaeger): no per-data licensing, full control, but you operate, scale, and maintain it — real engineering time. Buy a commercial platform (New Relic, Datadog, Dynatrace): fast to stand up, integrated, supported, but you pay — and the pricing is the whole story. New Relic’s model has centered on data ingest plus per-user pricing; competitors like Datadog price per host and per feature. Either way, the bill scales with how much telemetry you emit, which leads directly to the trap.

Where the cost bites

Commercial observability is notorious for bill shock, and the mechanism is worth understanding before adopting. Costs scale with data volume (every log line and metric ingested), cardinality (unique tag/attribute combinations — the same explosion that hurts Prometheus, but here it hits your invoice), and retention. Teams that instrument everything verbosely — debug logs, high-cardinality custom attributes, per-request traces at full sampling — can see costs balloon far beyond expectation. Disciplined telemetry (sampling traces, filtering noisy logs, bounding custom attributes) is what keeps a commercial platform economical, and is exactly the discipline teams skip until the first big invoice. The buy-vs-build math should account for this ongoing, volume-driven cost, not just the sticker price.

What people get wrong

• Ignoring volume-driven pricing — instrumenting everything at full fidelity produces bill shock; sampling and filtering are cost controls, not afterthoughts.
• Assuming buy is always cheaper than build: at large telemetry volumes, self-hosted open source can win despite the operational effort.
• High-cardinality attributes — unbounded custom tags inflate both performance and cost, just as they do in open-source metrics systems.

Primary source: New Relic documentation