The observability market is booming. Grafana Cloud wins deals, keeps customers, and creates long-term partner value. Here's everything you need to know.
Grafana Labs: 2× Gartner MQ Leader, furthest in Completeness of Vision (2025). Trusted by Anthropic, Bloomberg, NVIDIA, Microsoft, Salesforce. ~1,000 engineers building the next wave.
Per-host/per-seat pricing punishes growth. Bills can reach ~30% of cloud provider spend. Datadog is the most common source of "bill shock."
Proprietary agents, formats, and query languages. Switching means re-instrumenting everything. The #2 reason customers leave Datadog and Dynatrace.
4–5 tools open during an incident. Each has its own login, query language, and alert system. New engineers have to learn all of them.
First 15–30 min spent gathering data, not diagnosing. Only 2–3 senior engineers know the system. When they leave, knowledge leaves too.
AI can return 200 OK while giving a wrong answer. Traditional monitoring tracks uptime — not output quality, hallucinations, or token costs.
Cutting retention, disabling features, sampling data to control bills. Paying more to see less. Important signals get lost.
Ingest all data into their proprietary platform. Per-host/per-seat pricing. Proprietary agents = lock-in. Switching means starting over.
Connects to 180+ data sources. Queries data where it lives. Usage-based pricing. Built on open source + OpenTelemetry. Everything is portable.
Migration tip: Alloy natively ingests Datadog trace/metric format. Run both platforms in parallel — no re-instrumentation, no cutover risk.
Shared Slack channels with Grafana engineers, direct product team access, transparent pricing. "Datadog does not treat customers like partners. Grafana totally does." — CTO quote
Competitors' AI is conversational (helps write queries). Grafana Assistant is agentic — builds dashboards, investigates incidents, triages alerts. Plus: monitors your customers' own AI systems.
OTel-native. PromQL/LogQL are industry standards. Dashboards stored as JSON in Git. Alloy agent is open source. Zero proprietary lock-in anywhere.
APM, frontend, database, K8s monitoring, IRM, load testing, profiling, AI observability — all in one. "A fully furnished, move-in-ready house." Land-and-expand creates recurring partner growth.
Adaptive Telemetry removes data you pay for but never use across all 4 signal types. Savings: 30–50%. Lovable cut 93%. No competitor has this.
| Signal | Product | How It Works |
|---|---|---|
| Dashboards | Grafana | 180+ data sources. Multi-source alerting. Git Sync for dashboards-as-code. The world's most popular operational dashboard. |
| Metrics | Mimir | Prometheus-compatible. OTel, Graphite, Datadog, InfluxDB ingest. 13-month retention included. 10× query speed at scale. |
| Logs | Loki | Label-indexed (not full-text) = ~$0.55/GB vs. $1.50–$2.50+/GB at competitors. LogQL queries. Knowledge Graph integration. |
| Traces | Tempo | Distributed tracing across microservices. OTel, Jaeger, Zipkin. TraceQL for attribute-based trace queries. |
| Profiles | Pyroscope | Always-on continuous profiling, ~2–5% overhead. Line-by-line CPU/memory. AI flame graph analysis. $0.50/GB. No competitor equivalent. |
| Collection | Alloy | Open-source agent. OTel + Prometheus + Datadog formats. Fleet Management for centralized config at scale. |
| Testing | k6 + Synthetics | Load testing (k6) + proactive endpoint monitoring from global locations. Catch issues before users do. |
| Incidents | Grafana IRM | On-call, escalation, coordination, runbooks, post-mortems. Replaces PagerDuty/ServiceNow. $20/user vs. $21–$41 (PD) or $30 (DD). |
OTel-native APM. Auto-discovery, 100+ dashboards per entity, Knowledge Graph RCA. Alert → root cause in minutes.
Web Vitals, user actions, session replay. Correlates browser symptoms to backend traces. Built on Faro + Beyla.
Cluster health, cost attribution (OpenCost), efficiency views. Color-coded signals. Single Helm chart deployment.
Query-level detail: slow queries, execution plans, missing indexes. ~$50/instance vs. ~$85/host (competitors). MySQL + PostgreSQL.
| OSS | Enterprise | Cloud | |
|---|---|---|---|
| What | Free visualization + alerting (AGPLv3). Customer hosts everything. | Self-managed Grafana + RBAC, SAML, audit logging, enterprise plugins. | Fully managed platform: Grafana + Mimir/Loki/Tempo/Pyroscope + AI + solutions + IRM. |
| Infra | Customer operates, upgrades, scales | Customer operates; Grafana Labs provides license + support | Grafana Labs manages everything |
| Price | Free (+ compute/storage/1–2 FTEs) | Commit-based. 1 instance + 60 users base. | Usage-based. Free tier → Pro ($19/mo) → Enterprise ($25K/yr) |
| Security | Basic auth only | RBAC, SAML/LDAP, SCIM, audit logging | All Enterprise security included. SOC 2. 99.5–99.9% SLA. |
| AI | None | None (enterprise data source plugins only) | Full: Assistant, Knowledge Graph, Adaptive Telemetry, Investigations, Agent O11y |
Cloud includes everything in Enterprise plus more. Lead with Cloud in every conversation. Enterprise only when data physically cannot leave the customer's servers.
| Situation | Recommend | Why |
|---|---|---|
| Cloud-native or hybrid | Grafana Cloud | Full platform. No infra to manage. Alloy monitors on-prem + cloud. |
| Running Grafana OSS | Cloud (Migrate) | Same product. Zero retraining. Dashboards transfer. Warmest motion. |
| Large scale, own account | Cloud BYOC | Full Cloud in customer's AWS/GCP/Azure. 20%+ cheaper at scale. 99.9% SLA. |
| Hard on-prem-only mandate | Enterprise | Dashboard layer only. Customer runs own backends. Defense/classified use cases. |
| Regulated, cloud OK | Cloud Enterprise tier | SOC 2, custom SLAs, dedicated support. |
| Exploring, no budget | Cloud Free tier | 3 users, no credit card. Let the product sell itself. |
Cloud = the full house, furnished. Enterprise = the front door — they build the rest.
Prometheus, CloudWatch, Elasticsearch, MongoDB, BigQuery, Snowflake — query data where it lives. One dashboard across all sources.
Alert on error rate (Loki) AND CPU (Prometheus) in a single rule. No other platform does this natively.
Dashboards as JSON in Git. PR reviews, instant rollback, CI/CD deployment. Portable files the customer owns.
Auto-discovers services and dependencies from telemetry. Maps relationships. Runs 24/7 anomaly detection. No CMDB, no manual setup.
Every service, pod, node, database found from telemetry. Refreshes every minute. New deploys appear automatically.
5 structured categories: Saturation, Amend, Anomaly, Failure, Error. Generated automatically — no manual thresholds. No other platform has this.
Causality-driven investigation. Traverses up to 4 degrees of relationships. "Payment is slow → DB lock contention → recent deploy on order service."
Remembers past incidents and patterns. New engineers investigate guided by accumulated knowledge, not just raw data.
Impact: 30-min diagnosis → 3-min guided investigation. Pricing: No separate SKU. ~5–10% incremental on metrics cost.
Conversational. Helps write queries. You still do the work. Confined to vendor's own data.
Agentic. Builds dashboards, investigates incidents, triages alerts, drafts fixes via GitHub/GitLab. Works across 180+ data sources.
| Product | Status | What It Does |
|---|---|---|
| Assistant | GA | Natural language queries, dashboard building, system explanation, MCP integrations for external tools. |
| Investigations | GA | Parallel AI analysis across metrics/logs/traces/profiles. Root cause 3.5× faster (8 min vs. 28 min). |
| Alert Enrichment | GA | Auto-adds context (service relationships, deploy history, past incidents) before anyone looks at the alert. |
| Automations | GA | Scheduled workflows: health checks, capacity reports, recurring alert investigation. |
| gcx + MCP | GA | CLI tool + MCP server. Connect Claude Code, Cursor, Copilot directly to Grafana Cloud data. |
| Watchers | Preview | Proactive AI agents detecting slow degradation humans would miss. |
| o11y-bench | GA | Open-source benchmark for AI observability agents. No competitor equivalent. |
The problem: Healthy ≠ correct. A service returning 200 OK can give a wrong, harmful answer. Agent O11y closes this gap.
Every prompt, response, tool call, token count, and cost — end to end. Debug bad answers by replaying the conversation. SDKs: Go, Python, TS, .NET, Java.
Token spend by project, agent, team, model. Test expensive models vs. cheaper alternatives. Chargeback with evidence.
Check prompts before they reach the LLM. Block PII, API keys, sensitive data. Verify tool calls match allowed patterns.
Quality scoring on live production traffic. Flag drift before customers notice. LLM-as-judge + human review.
Frameworks: LangChain, OpenAI Agents, LlamaIndex, Google ADK, Strands, LiteLLM. Plugins: Claude Code, Cursor, Codex, Copilot CLI. Net-new market — competitors have basic support at best.
| Product | Status | How It Works |
|---|---|---|
| Adaptive Metrics | GA | Finds unqueried metrics, recommends aggregation/drops. Avg savings: 30–50%. SimpliSafe: 100 series → 10 (90% cut). |
| Adaptive Logs | GA | Identifies never-queried log patterns (debug, health checks). Recommends drop/sample rules. Preserves error logs. |
| Adaptive Traces | GA | Managed tail-based sampling. Keeps errors + high-latency at full fidelity. Samples routine traces. No manual pipeline config. |
| Adaptive Profiles | GA | Lower resolution normally, full fidelity during deploys/anomalies/investigations. Event-aware, not static. |
Process: Analyze usage (14–30 days) → Generate recommendations with savings estimate → Review in UI or CLI → Apply with a click → Continuous iteration.
No other vendor reduces costs automatically across all 4 signal types.
"Adaptive Telemetry cut our metrics costs enough to fund logs and traces. All three pillars for less than metrics alone used to cost."
6 vendors → Grafana Cloud. 2K+ resources incl. AI agents. Assistant Investigations found RAG pod memory issue in minutes. Use when: multi-tool + AI workloads.
ObsDeck on Grafana: step-by-step troubleshooting cut incident engagement from 60 min to under 1 min. Use when: complex microservices, high MTTR.
OSS → Cloud. $640K+ saved. 2 FTEs freed. 40% MTTR improvement. Use when: self-hosted Grafana/Prometheus.
NR cost for one division = entire Grafana spend across all systems. Use when: NR displacement, per-seat frustration.
Replaced 4 tools. Same or lower cost. "Stone age before." Use when: tool consolidation.
Tamed 1M+ series with Adaptive Metrics. Monitors LLM tokens by model/region. Use when: DT displacement + AI.
All 3 pillars for less than metrics alone used to cost. Use when: expand coverage without more budget.
Most dramatic published Adaptive Telemetry result. Use when: high-cardinality concerns or growth fears.
#1 displacement trigger: unpredictable, escalating costs. #2: vendor lock-in.
| Area | Grafana | Datadog |
|---|---|---|
| Pricing | Usage-based. Adaptive Telemetry saves 30–50%. | Per-host + per-feature. Bills reach ~30% of cloud spend. |
| Architecture | Open. 180+ sources. OTel-native. Portable. | Polished UX. But proprietary agent, forced ingestion, lock-in. |
| AI | Agentic — takes action across any data source. | Bits AI: conversational. Confined to DD platform. |
| Logs | ~$0.55/GB | $1.50–$2.50+/GB. Better full-text search OOTB. |
| Profiling | eBPF. $0.50/GB. Always-on. | $19/host + $2/container. Manual instrumentation. |
| IRM | $20/user. Full suite. | $30/seat on top of platform. |
| Migration | Alloy ingests DD format. Parallel run. No re-instrumentation. | No comparable exit path. |
| Area | Grafana | New Relic |
|---|---|---|
| Pricing | Usage-based. No per-seat fees. Unlimited users. | $49–$419/seat + data credits. Per-seat limits observability access. |
| Retention | 13-month metrics default. No tiered fees. | 8-day default. Extended retention costs extra. |
| Dashboards | Preferred by practitioners. Multi-source. Git Sync. | Pre-built OK. Custom hits limits. |
| Open Standards | OTel-native. PromQL/LogQL are industry standards. | Claims OTel but leans proprietary (NRQL). |
| AI | Agentic + Knowledge Graph + Agent O11y. | Competitive query assistance. No SAAFE, no agentic action. |
Narrative: "Big Tent vs. Walled Garden." Be honest about DT strengths — credibility wins.
| Area | Grafana's Edge | Dynatrace's Gap |
|---|---|---|
| AI | Agentic — builds, investigates, triages | Davis: chat copilot only |
| Architecture | 180+ sources. Query in place. | Must ingest into Grail. |
| AI O11y | First-party SDKs. Production evals. | Basic. Third-party SDK. |
| Cost | Usage-based. Adaptive Telemetry. | Fixed commit. No auto reduction. |
| IRM | Full suite: OnCall + Incidents + runbooks. | Alerting only. Needs PagerDuty/ServiceNow. |
| Community | ~10K active. Transferable OTel skills. | Small. Proprietary DQL. |
Theirs: 10-year retention, one query lang. Ours: ~13-month. Reframe: most teams query last 7–30 days. DQL is proprietary.
Theirs: Production breakpoints. Counter: Pyroscope = always-on profiling fleet-wide. DT has no equivalent.
Theirs: Single binary, day-one coverage. Reframe: Simplicity = lock-in. Grafana is OTel-first. Portable.
Theirs: Runtime security. Ours: No AppSec. Qualify early. Most customers have dedicated security tools.
Your Grafana Channel Manager is your first point of contact for deal support, co-selling, enablement, and customer introductions.