Azure Monitor vs Datadog: App Insights Path, Free Metrics Cliff, Multi-Cloud

Azure Monitor and Datadog both claim "cloud monitoring." On Azure they can look interchangeable for a week: charts appear, alerts fire, someone demos a dashboard. The products diverge on who owns the control plane and what stops being free.

Azure Monitor is the native stack for Azure resources: platform metrics, activity logs, Log Analytics (KQL), and Application Insights for APM and availability tests. Telemetry that matters for apps usually lands in a Log Analytics workspace and rides the Azure invoice.

Datadog is a multi-cloud vendor control plane. You install an agent (or Operator), enable integrations, and pay primarily per host and per module whether the workloads sit in Azure, AWS, GCP, or a mix.

The spine of this page is Azure-native App Insights + Log Analytics vs multi-cloud Datadog, plus a hard look at when "free" Azure metrics stop being the real story. It is not the AWS-native duel (Datadog vs CloudWatch), not PromQL vs turnkey (Grafana Cloud vs Datadog), not per-host vs GB+seats (Datadog vs New Relic), not Dynatrace DPS (Datadog vs Dynatrace), and not the nine-tool category map (best cloud monitoring tools).

List rates below follow public Azure Monitor and Datadog pages as of July 2026. Regions, EA discounts, and commitment tiers change the dollars. The shape of the cliff does not.

TL;DR

Decision spine: App Insights path vs multi-cloud pane

On Azure, the productive path for application health is usually:

1. Platform metrics for VMs, App Service, AKS, SQL, Load Balancer (free standard metrics).
2. Diagnostic settings that stream resource logs into a Log Analytics workspace.
3. Application Insights (workspace-based) for requests, dependencies, exceptions, and availability tests.
4. Alerts and workbooks on top of that workspace.

That path is excellent when every production dependency is Azure-shaped and on-call already thinks in KQL and Azure Resource Graph. It is a silo the moment you need the same service map for an AWS account, a GCP project, or a datacenter cluster beside AKS.

Datadog's path is agent-first and cloud-agnostic. The Azure integration can pull metrics and resource metadata, but the product thesis is still "one vendor graph across environments," not "deepest native Azure console experience." You pay for that with a second MSA, agent lifecycle, and a host-metered floor.

Pick Azure Monitor when Azure is the estate. Pick Datadog when Azure is one region of the estate. Pretending App Insights is a multi-cloud APM is how teams bolt on a second tool six months later under incident pressure.

When "free" Azure metrics stop

Marketing language around Azure Monitor often collapses into "metrics are free." That is true for a narrow slice and misleading for the bill that actually shows up.

What stays free (important):

• Standard platform metrics for Azure resources are ingested at no charge (unlimited per Microsoft's platform metrics pricing). CPU on a VMSS, App Service request rates, many AKS/node charts in Metrics Explorer: this layer really is free.
• Activity log retention for the default window and a limited set of activity-log alert rules.
• A small free allotment on metric alert time-series (commonly cited as 10 monitored metric time-series per month on the public pricing table; beyond that you pay).
• Application Insights URL ping web tests are free. Telemetry those tests generate can still count as workspace data depending on configuration.

What is not free (the cliff):

1. Log Analytics Analytics Logs ingestion. Pay-as-you-go Analytics Logs include 5 GB/month free per billing account. Everything interesting after "glance at Metrics Explorer" tends to land here: resource diagnostic logs, Container Insights, App Insights request/dependency/exception tables, custom logs. Regional list rates are often roughly $2+/GB before EA discounts; commitment tiers (from 100 GB/day) lower effective per-GB price if you can commit.
2. Application Insights itself. Workspace-based App Insights stores in Log Analytics. There is no separate magic free APM lake. Instrument the app, ship traces and logs, and the workspace meter moves. Classic App Insights resources follow similar pay-as-you-go ingest without commitment-tier flexibility.
3. Retention beyond included windows. Analytics Logs include a short interactive retention window (on the order of 31 days, longer in some Sentinel-enabled setups). App Insights tables commonly include 90 days without extra retention charge. Keep years of KQL-hot data and you pay retention.
4. Diagnostic settings turned wide open. Sending every category from every resource to Analytics Logs is the classic Azure bill shock. The metrics blade still looks "free" while the workspace doubles weekly.
5. Standard (and multi-step) web tests. Ping tests are free; standard web tests charge per scheduled execution. Multi-step tests bill per active test per month. Alert rules and notifications are separate meters.
6. Azure Monitor workspace / Managed Prometheus samples. Metrics you send into an Azure Monitor workspace (including Managed Prometheus) bill by samples ingested, not by the free platform-metrics path.

Rule of thumb: if your only SLIs are platform metrics in Metrics Explorer with a handful of metric alerts, Azure Monitor can stay cheap or free on the monitoring line. The day you need App Insights dependency maps, Container Insights, and broad diagnostics, you are in Log Analytics economics. That is when Datadog's host spreadsheet starts to look comparable, not because Azure "got expensive overnight," but because you left the free metrics layer.

How you operate it: Portal + KQL vs Agent + modules

Azure Monitor is owned by whoever already owns the subscription. Config is Portal, ARM, Bicep, or Terraform azurerm_monitor_* / App Insights resources. Access is Entra ID and Azure RBAC. Debugging starts in Metrics, Logs (KQL), App Insights Application Map, and Workbooks. There is an Azure Monitor Agent when you need guest OS or custom logs, but basic Azure resource metrics do not require a third-party agent.

Datadog is owned by whoever owns the agent rollout and the Datadog org. You install the Agent or Kubernetes Operator, wire the Azure integration, and usually enable APM libraries or admission controllers separately. Config spreads across agent YAML, the Datadog UI, and the Datadog Terraform provider. The payoff is consistent tagging and one query surface across clouds. The cost is a second system that must stay healthy: keys, sites, upgrades, and custom-metric cardinality.

Staff the model you already have. An Azure platform team fluent in Bicep and KQL should not invent Datadog for a single subscription. A company that already standardizes on Datadog for AWS should not invent a second pager culture for Azure "because metrics are free."

Feature depth only where this pair splits

Application Insights vs Datadog APM

App Insights gives request/dependency tracking, failures, performance, and an application map for instrumented .NET, Java, Node, Python, and other supported stacks. Auto-instrumentation on App Service and some container scenarios reduces SDK toil. Data gravity is the Log Analytics workspace: same KQL skills, same retention and ingest knobs.

Datadog APM is a separately priced host module on list rates (~$31/host/month annual on top of Infrastructure). Correlation into infra hosts, logs, and synthetics is the product pitch. You do not inherit Azure Portal muscle memory; you inherit Datadog's.

Pick App Insights when the apps and the on-call culture are Azure-native. Pick Datadog APM when traces must sit next to non-Azure services in one UI, or when the org already bought the Datadog stack and refuses a second APM.

KQL workspaces vs Datadog log pipelines

Log Analytics is the cost and skill center of serious Azure Monitor usage. Table plans (Analytics / Basic / Auxiliary), transformation rules, and workspace design decide both query power and invoice shape. Datadog separates ingest, index, and retention; you can cheap-ingest and selectively index. Azure burns you when diagnostics and App Insights share one hot Analytics table plan without filters. Datadog burns you when you index everything at long retention.

Availability tests vs Datadog Synthetics

App Insights URL ping tests are free and fine for "is this URL up from Azure." Standard web tests add multi-location scripting at a per-execution price. Datadog Synthetics (API ~$5/10k runs, browser ~$12/1k runs annual list) is stronger when product/QA owns journeys and already lives in Datadog. Do not migrate ping tests to Datadog browser checks without a reason; do not pretend free ping tests replace multi-step checkout coverage.

Worked sketch: same Azure fleet, two invoices

Fleet: 40 Azure VMs / node-equivalent hosts, App Insights on the main API, 180 GB/month into Log Analytics (diagnostics + App Insights + some container logs), 8,000 availability/synthetic API runs/month. No AWS/GCP yet.

Azure Monitor path (illustrative)

Turn on verbose diagnostics for every storage account and the 180 GB becomes 500 GB. The free metrics blade does not move. The workspace does.

Datadog path (illustrative, annual list)

Read the duel correctly: for an Azure-only estate that needs App Insights-depth telemetry, Azure Monitor often wins on list-rate dollars if you govern workspace ingest. Datadog's floor is the host × APM stack, which is rational when multi-cloud or org-standard agents matter more than the Azure invoice line. Skipping APM on Datadog (infra only ~$600) narrows the gap but also drops the trace story App Insights was giving you.

Pick Azure Monitor if / Pick Datadog if

Pick Azure Monitor if:

1. Production is Azure-only (or Azure plus a thin edge) and on-call already uses Portal + KQL.
2. Application Insights covers the languages and hosting you run (App Service, Functions, AKS with supported instrumentation).
3. You can name an owner for Log Analytics ingest (table plans, diagnostic settings, retention).
4. Finance wants observability on the Azure bill, not a second SaaS commitment.
5. Free platform metrics plus disciplined workspace usage beat a per-host APM line.

Pick Datadog if:

1. You run Azure beside AWS/GCP/on-prem and need one service graph.
2. The company already standardized on the Datadog Agent and Terraform provider.
3. Product/QA owns browser synthetics next to APM and refuses a second tool.
4. You will not staff KQL + workspace cost governance.
5. Multi-cloud tagging and Watchdog-style feeds matter more than native Azure Insights blades.

Stay mixed if: Azure Monitor keeps platform metrics, activity logs, and Azure-native alert actions (autoscale, Azure functions), while Datadog owns a subset of customer-facing APM across clouds. Give each a written scope. Eternal dual-shipping of every log line is how both invoices win.

Migration path (this pair)

Stay on Azure Monitor. Cap diagnostic settings. Move noisy tables to Basic/Auxiliary where query needs allow. Set a workspace daily cap or alert on ingestion. Keep App Insights workspace-based so commitment tiers remain available. Prefer ping tests until a journey truly needs standard web tests.

Add Datadog on top. Enable the Azure integration, start Infrastructure on a slice of hosts, keep Azure metric alerts that drive Azure automation. Do not mirror every Log Analytics table into Datadog logs on day one.

Switch primary to Datadog. Roll Agent/Operator, map the pages you page on (not every workbook), replace App Insights for services that must join the multi-cloud map, then reduce diagnostic duplication so Azure is not still ingesting a full copy. Budget a quarter.

Switch back toward Azure Monitor. Usually after Datadog bill shock on an Azure-heavy estate. Keep Datadog only where multi-cloud traces still earn their keep; move Azure-only services back to App Insights; cut custom metrics and unused integrations during overlap.

Failure modes

Azure Monitor: believing "metrics are free" after enabling App Insights + Container Insights + blanket diagnostics; one giant Analytics workspace with no table-plan strategy; KQL alerts that scan huge ranges; Sentinel on the same workspace as noisy app logs (security meters on operational data); ping tests mistaken for journey coverage.

Datadog: host and APM creep on node pools; custom-metric cardinality; indexing every Azure log copy you also store in Log Analytics; treating the Azure integration as "free CloudWatch-style metrics" while still paying Datadog hosts.

Shared: neither stack alone answers "checkout is down because a vendor API outside Azure degraded." Platform metrics and APM see your side. External dependency correlation is a different layer.

Bottom line

Azure Monitor vs Datadog turns on whether you want the App Insights + Log Analytics + KQL native path on a single cloud, or a multi-cloud agent platform with a host-metered floor. Free Azure platform metrics are real. They stop being the bill the moment workspace ingest becomes your source of truth.

If users can be down while both Azure workbooks and Datadog are green, add black-box monitoring after you pick a meter. DevHelm covers that external layer. For the wider category, see best cloud monitoring tools. For the AWS-native twin of this debate, see Datadog vs CloudWatch.