SigNoz vs Datadog — The Open-Source APM Challenger

Our Datadog bill crossed a number I don't want to print, and the finance team started asking questions I couldn't answer well. Most of the spend wasn't APM. It was custom metrics we'd forgotten about, indexed logs nobody queried, and a per-host count that grew every time an autoscaler had a busy afternoon. So we ran a real evaluation of SigNoz next to it, in staging, for six weeks. This is what we learned.

Both tools answer the same question: what is my system doing right now, and why did it break at 3 a.m. The difference is how much they charge to answer it, and how much you own the answer once you have it.

What SigNoz actually is#

SigNoz is an open-source observability platform. Traces, metrics, and logs live in one place, backed by ClickHouse, and it speaks OpenTelemetry as its native protocol rather than as a bolted-on import path. You can run it yourself on your own hardware, or pay for SigNoz Cloud and let them run the ClickHouse cluster.

That OTel-first design is the part worth sitting with. You instrument once with the OpenTelemetry SDK, point the collector at SigNoz, and you're done. No proprietary agent, no vendor-specific tracer buried in your app. If you ever swap SigNoz out, your instrumentation stays. Datadog supports OTel too, but its center of gravity is still the Datadog Agent and its own libraries, and you feel that pull the moment you want a feature the OTel path doesn't cover.

Cost, which is why most people read this#

Datadog prices per host, per million spans, per GB of logs ingested and again per GB indexed, and separately for custom metrics. Each line is reasonable on its own. Added together across a real fleet, they compound into a bill that grows faster than your traffic does. The custom-metrics charge is the one that ambushes people, because a single high-cardinality tag can multiply your metric count without anyone noticing until the invoice lands.

SigNoz self-hosted has no per-host license. You pay for the machines ClickHouse runs on and the engineer-hours to keep it healthy. For high-volume telemetry that inverts the economics: ingest more and your marginal cost is disk and CPU, not a metered rate. SigNoz Cloud sits in the middle — usage-based pricing on ingested data, no per-host tax, and no separate charge to index logs you already sent. For most teams we've modeled, SigNoz Cloud lands well under Datadog for the same volume, and self-hosting beats both on raw cost if your telemetry is large enough to justify the operational work.

That "if" is the whole argument.

The self-hosting tax nobody quotes you#

Running SigNoz yourself means running ClickHouse yourself. ClickHouse is fast and it is not free to operate. You own retention and TTLs, disk sizing, schema migrations across upgrades, backups, and the 2 a.m. page when a node fills its volume during a traffic spike. At small scale a single box is fine. At the scale where the cost savings get exciting, you're operating a stateful database that is now itself a production system needing its own monitoring.

Datadog's pitch is the exact opposite: you run nothing. No cluster, no storage math, no upgrade windows. For a team without a spare platform engineer, that zero-ops promise is worth real money, and pretending otherwise to save on a subscription is how you end up with two problems instead of one.

Feature breadth#

Datadog is broader, and it's not close. Synthetics, RUM, database monitoring, network performance, CI visibility, cloud security, a sprawling catalog of integrations, and a maturing set of AI features. If your team lives across a dozen of those surfaces, replacing Datadog means replacing a dozen tools, not one.

SigNoz covers the core exceptionally well: distributed tracing, metrics dashboards, log querying, exceptions, and alerting, all correlated through the same OTel data. For a lot of teams that core is 90 percent of what they open Datadog for anyway. The honest read is that SigNoz competes hard on APM, logs, and metrics, and doesn't try to be a security or RUM platform. Know which columns you actually use before you assume you need the wide one. Our own audit of APM tools made that painfully clear.

The comparison#

Dimension	SigNoz	Datadog
Model	Open-source, self-host or Cloud	Proprietary SaaS
Pricing	Infra cost (self-host) or usage-based Cloud, no per-host fee	Per-host, per-GB, per-metric, indexed separately
Native protocol	OpenTelemetry	Datadog Agent (OTel supported)
Data store	ClickHouse (yours or theirs)	Datadog's backend
Data ownership	Full, on self-host	Vendor holds it
Feature breadth	Core APM, logs, metrics	Very broad platform
Ops burden	You run ClickHouse (self-host)	None
Maturity/ecosystem	Younger, growing fast	Deep, established

Data ownership and lock-in#

Self-hosted SigNoz keeps your telemetry inside your own boundary. For teams in regulated industries, or anyone nervous about shipping request payloads to a third party, that's not a nice-to-have. Combined with OTel instrumentation, it means neither your data nor your data collection is hostage to a vendor's roadmap or pricing changes.

Datadog owns your data by design. That's the deal you accept for zero-ops, and it's a fine deal right up until a renewal negotiation or a compliance review makes you wish it weren't.

Maturity#

Datadog has years of hardening, a mature alerting engine, and an integration for nearly everything you run. SigNoz is younger. It's improved quickly and the OTel foundation means it inherits the whole ecosystem's momentum, but you will occasionally hit a rough edge Datadog polished off years ago. Weigh that against a bill you can actually predict.

The call we'd make#

If your team is OTel-first, cost-conscious, and has the engineering muscle to run a ClickHouse cluster, or you'd rather pay for SigNoz Cloud than Datadog's metered everything, go SigNoz. You get the core observability story, real data ownership, and a bill that tracks your telemetry instead of your host count. If you need breadth across synthetics, RUM, and security, or you have nobody to babysit a database, Datadog earns its price and you should stop agonizing. We split the difference: SigNoz for the high-volume services where we control the stack, Datadog for the corners where its breadth still wins. Pick for the columns you use, not the ones on the pricing page.