Datadog

What Datadog actually is

Datadog is the reference point for enterprise observability: infrastructure monitoring, APM, log management, synthetic monitoring, real user monitoring, security and compliance products, all fed into one platform with one agent and one set of dashboards. It has been around since 2010, when Olivier Pomel and Alexis Lê-Quôc founded the company in New York City; infrastructure monitoring shipped in 2012, the company listed on Nasdaq in 2019, and it now employs several thousand people with the headquarters still in the New York Times Building. For most engineering teams evaluating monitoring in 2026, Datadog is the default against which everything else is compared — including the open-source stacks in /category/monitoring-hosting that try to undercut it.

The product is genuinely broad. One agent collects metrics, traces and logs; the platform correlates them, so a latency spike on a service map can take you to the offending trace, the host's CPU graph, and the deploy that caused it. The integration catalogue is the deepest in the market — the company and third-party listings variously claim 600 to over 1,000 integrations — and the query language, alerting, and dashboard tooling are mature enough that most teams never hit a wall in the product itself. What they hit instead is the bill.

Pricing: what you really pay

Datadog's list pricing is public and per-host, which is more transparent than the enterprise-observability norm. The free tier covers 5 hosts with one day of metric retention — enough to evaluate the platform, not enough to run anything. Infrastructure Pro is $15 per host per month on annual billing ($18 on demand), Infrastructure Enterprise $23 ($27 on demand). APM is priced per host as a separate product: $31 per host per month when bundled with Infrastructure on an annual contract, $36 standalone, rising to $40-plus on Enterprise tiers. Logs are metered twice — $0.10 per GB ingested plus $1.70 per million events indexed for 15-day retention — and there are separate meters for everything else: synthetic API tests at $5 per 10,000 runs, browser tests at $12 per 1,000 runs, RUM at $1.50 per 1,000 sessions, Cloud Security Posture Management at $7.50 per host, and overage on indexed spans at $1.70 per million.

Where teams get hurt is stacking. A ten-host startup running Infrastructure Pro, APM on five hosts and modest logs lands around $360 per month at list. A 100-host team with APM everywhere, 30 GB of logs per day and some custom-metric overage is a $5,000-plus monthly bill — before negotiated discounts. Containers complicate it further: Pro includes 10 containers per host (20 on Enterprise) and bills beyond that at a per-container-hour rate. APM includes one million indexed spans per host per month, and a busy API service can burn five to ten million, adding a fifth to a half on top of the base rate. Our practical advice: treat the published per-host price as the floor, model your span and log volumes before signing, and negotiate — mid-market discounts of 20 to 30% are common enough that list price is a starting point rather than the answer.

One billing detail worth understanding before you negotiate: the difference between Pro and Enterprise is not just machine-learning features. SAML SSO, advanced role-based access control, governance tooling and the longer retention options sit behind Enterprise, so compliance requirements can force the $23-per-host tier faster than feature needs do. Teams below roughly 50 hosts generally find Pro's alerting sufficient; teams above 200 hosts tend to want Watchdog simply because nobody can maintain that many alert rules by hand. Pricing rises with headcount in both senses, and that is worth modelling at your actual host count rather than your current one — most organisations arrive at Datadog with 20 hosts and stay until they have 300.

Performance & infrastructure

Datadog is a SaaS platform rather than a hosting provider, so "infrastructure" means where your telemetry lives. The platform runs as independent regional sites chosen at sign-up: US1, US3 and US5 in the United States, EU1 in Germany, UK1 in the United Kingdom, AP1 in Japan, AP2 in Australia, and FedRAMP High-certified US government sites. Sites are fully isolated — data does not cross them, and moving an organisation between sites is a rebuild rather than a setting — which makes the choice at sign-up a real architecture decision for anyone with data-residency obligations. The EU site in Germany is the answer for GDPR-sensitive European workloads.

On capability, the interesting engineering is in scale and correlation. The agent is lightweight and well maintained; ingestion handles metrics, traces and logs from essentially anything; and the platform's value is in joins — traces linked to logs linked to infrastructure state linked to deploy events. Retention on paid plans runs to 15 months for metrics, with Flex Storage options for cheaper long-tail log retention and log archives to your own S3 or GCS bucket. The Enterprise tier adds machine-learning alerting (Watchdog), Live Processes, governance tooling, SAML SSO and advanced RBAC — the features that large organisations usually discover they need a contract renewal too late.

We have not benchmarked Datadog's ingestion pipeline ourselves — our /benchmarks programme tests compute, not observability SaaS — so treat throughput and query-speed statements here as vendor and community documentation, not measured results.

Where it falls short

The billing model is the product's biggest weakness, and it is structural rather than accidental. Datadog charges per host, per product, per usage meter, and the meters that surprise people are logs and indexed spans — the two things that grow fastest in production. Custom metrics above the included quota (100 per host on Pro, 200 on Enterprise) are billed as overage. Engineering leaders consistently report that forecasting the Datadog bill is harder than forecasting infrastructure spend, and that the bill's growth rate exceeds their traffic's growth rate. The company's financial results suggest we are not the only ones to notice.

Second, the free tier is evaluation-grade. One day of retention and five hosts means you cannot run a real side project on it the way you can with some competitors' free tiers. Third, product sprawl is a governance problem: security monitoring, RUM, synthetics, CSM and the rest are excellent individually but multiply the SKU count, and each SKU has its own price list. Fourth, cost control tooling exists but is bolted on — budget alerts and usage APIs help, yet the default trajectory of a Datadog deployment is upward, and discipline is on you. Finally, while the docs are strong, the sheer surface area means onboarding a new engineer into your Datadog setup is a project of its own.

The competitive frame is worth stating plainly, because Datadog's weakest dimension is price and that is where its rivals attack. Grafana Cloud and the open-source stack win on cost for teams willing to own the plumbing. New Relic's per-user pricing is easier to forecast and can be cheaper at moderate host counts. The hyperscalers' own monitoring tools win by default for teams that want one bill. Datadog's answer to all three is depth of correlation and breadth of integrations, and for the right team that answer is correct — but it is an answer you should price against those alternatives with your own numbers, not with a sales deck.

Who should pick this

Our pick for scaling engineering organisations that need full-stack correlation and can pay for it: teams running microservices where the trace-to-log-to-metric workflow saves real debugging time, companies with compliance requirements that value the FedRAMP High government sites and Enterprise RBAC, and anyone whose incident response depends on one place where everything is visible. If your business loses money during downtime, the platform's ability to shorten incidents is the argument that matters, and it is a good one.

We would not point small teams, bootstrapped products, or cost-sensitive workloads at Datadog without first pricing the open-source alternatives seriously. A Prometheus/Grafana/Loki stack or a leaner commercial competitor can cover the same ground at a fraction of the cost if someone on the team will own it. If you do choose Datadog, our recommendations are concrete: commit annual for the discount, pick your site deliberately, cap custom metrics at the source, index a deliberate fraction of logs, and set budget alerts on every meter before the first invoice rather than after.

Key takeaways

• Datadog is the enterprise observability standard: one agent, one platform across infrastructure, APM, logs, synthetics, RUM and security, in market since 2010.
• Public per-host pricing: free for 5 hosts with one day of retention, Infrastructure Pro at $15/host/mo annual, Enterprise at $23, APM from $31/host when bundled.
• Usage meters stack on top — $0.10/GB log ingest, $1.70/million indexed events, span and custom-metric overages — and they are what push real bills past forecast.
• Data residency is decided at sign-up: fully isolated sites in the US, Germany, the UK, Japan, Australia, plus FedRAMP High US government sites.
• The free tier is for evaluation only; expect negotiated annual contracts and active meter discipline to keep costs sane in production.
• Best value for scaling teams that monetise reliability; poor fit for small projects, where open-source stacks in /category/monitoring-hosting win on cost.

Sources

• https://en.wikipedia.org/wiki/Datadog
• https://docs.datadoghq.com/getting_started/site
• https://stackcompare.net/datadog-pricing-2026-infrastructure-apm-logs-what-you-will-really-pay
• https://stackscored.com/pricing/observability/datadog
• https://apis.io/plans/datadog/datadog-plans-pricing
• https://procr.io/post/datadog-pricing-2026
• https://getpricepulse.com/companies/datadog-pricing.html
• https://dpa-atlas.foundagent.net/gdpr/datadog