How much does cloud monitoring cost?

Cloud monitoring ranges from $0 (open-source self-hosted, free vendor tiers) to $200,000+ per month for enterprise deployments. The median mid-market team running 100 hosts with full observability spends $3,000 to $15,000 per month. Costs are driven primarily by host count, log volume, custom metric cardinality, and APM coverage. Verify current pricing on each vendor's pricing page before purchasing.

What is the cheapest cloud monitoring tool?

Self-hosted Prometheus and Grafana is the cheapest in pure license terms (free), but requires platform engineering capacity. Among managed services, Grafana Cloud has the most generous free tier and lowest paid-tier pricing. New Relic's 100 GB free data ingest is the next most generous. Datadog and Dynatrace are positioned at the higher end of the market.

What percentage of cloud spend goes to monitoring?

Industry research from independent and vendor sources puts the median at 7 to 8 percent of cloud infrastructure budget. Below 3 percent typically signals under-monitoring. Above 15 percent signals overspend and is a common trigger for cost-reduction projects. Microservices and Kubernetes-heavy stacks land at the higher end of the range.

Why is my Datadog bill so high?

Bill spikes usually trace to one of: custom metrics generated from high-cardinality labels (Kubernetes pod, container, namespace), log indexing charges on top of ingest charges, APM span indexing above the included 1M per host, retention upgrades beyond the 15-day default, or container counting that bills each pod as a fractional host. See vendor pricing pages and our hidden costs page for the underlying mechanics.

Optimise

Twelve ways to cut your monitoring bill by 30 to 50 percent

Q: How can I reduce monitoring costs?

The highest-impact strategies in order: filter and sample logs at the source (saves 30 to 50 percent), control custom metric cardinality (saves 20 to 40 percent), sample APM traces at 5 to 10 percent (saves 15 to 30 percent), right-size retention to actual query patterns (saves 10 to 20 percent), and consolidate or migrate to open source where the team has capacity (saves 40 to 90 percent).

Verified April 2026

Twelve strategies, ranked by saving potential and implementation effort. Vendor-neutral, with implementation notes for the major platforms.

TL;DR

96 percent of organisations are actively cutting observability costs. The median team overspends by 30 to 60 percent. The four highest-impact strategies (log sampling, metric cardinality, APM sampling, retention tiering) typically combine for 35 to 55 percent saving inside a single quarter, before changing vendor.

Strategy matrix

#	Strategy	Typical saving	Effort
01	Filter and sample logs at the source	30 to 50 percent	Low
02	Cap custom metric cardinality	20 to 40 percent	Medium
03	Sample APM traces at 5 to 10 percent	15 to 30 percent	Low
04	Right-size retention	10 to 20 percent	Low
05	Tier hot, warm, and cold storage	30 to 60 percent on log retention	Medium
06	Negotiate annual commitment	15 to 25 percent off list	Low
07	Move dev and staging to a free tier	10 to 20 percent	Low
08	Consolidate overlapping vendors	15 to 30 percent	Medium to High
09	Migrate metrics to open source	60 to 90 percent on metrics line	High
10	Use Grafana Cloud as a managed open-source bridge	40 to 70 percent vs Datadog	Medium
11	Adopt OpenTelemetry from day one	Avoids future migration cost	Medium
12	Audit quarterly	Sustains all of the above	Low

Twelve strategies in detail

Filter and sample logs at the source

saves 30 to 50 percent

Logs are typically 50 percent of total observability spend. Drop health-check, framework, and load-balancer noise at the agent (Fluent Bit, Vector, Filebeat). Sample DEBUG and INFO at 10 to 20 percent while keeping all WARN and ERROR. Highest single lever in the toolkit.

Cap custom metric cardinality

saves 20 to 40 percent

Audit the top 10 highest-cardinality metric series. Drop user_id, request_id, IP address from metric labels (keep them in logs and traces). Convert per-URL gauges to bucketed histograms. Use Datadog Metrics Without Limits or equivalent aggregation rules to enforce caps.

Sample APM traces at 5 to 10 percent

saves 15 to 30 percent

Head-based sampling for high-volume services, tail-based sampling for error-relevant traces. 100 percent tracing is rarely necessary. Most teams discover the gap in fidelity is invisible at 10 percent and saves a meaningful share of the APM line.

Right-size retention

saves 10 to 20 percent

Default to 15 days for hot data. Push 30 to 90 day historical data to object storage (S3, GCS) and rehydrate on demand. Audit compliance requirements: most regulations require specific log types (auth, audit) for fixed periods, not all logs.

Tier hot, warm, and cold storage

saves 30 to 60 percent on log retention

1 second resolution for 24 hours. 1 minute for 7 days. 5 minute for 30 days. Hourly aggregates for 13 months. Most operational analysis happens in the 7-day window. Capacity planning needs hourly granularity at most.

Negotiate annual commitment

saves 15 to 25 percent off list

Vendors discount 15 to 25 percent for an annual or multi-year commitment with a usage floor. Negotiate exit terms, true-up windows, and the floor before signing. Time renewal negotiations to coincide with quarter-end vendor pressure.

Move dev and staging to a free tier

saves 10 to 20 percent

Production observability rarely needs to apply to ephemeral dev environments. Run dev/staging on Grafana Cloud free tier or self-hosted Prometheus. Typically 30 to 40 percent of monitoring spend is non-production environments masquerading as production.

Consolidate overlapping vendors

saves 15 to 30 percent

Datadog plus PagerDuty plus Splunk plus Sentry plus a homegrown dashboard. List every paid signal source. Eliminate any signal type covered by two or more platforms. The migration cost is real and quantified on the hidden costs page.

Migrate metrics to open source

saves 60 to 90 percent on metrics line

Self-host Prometheus and Grafana, pay only the underlying compute. Tempo for traces, Loki for logs. Most viable when there is a platform engineering function or strong DevOps culture. Quantified TCO comparison on the open-source-vs-paid page.

Use Grafana Cloud as a managed open-source bridge

saves 40 to 70 percent vs Datadog

Best transition point between fully self-hosted Prometheus and a fully commercial platform. Generous free tier, OpenTelemetry-native, no vendor lock at the data format level. Ideal for teams that want to leave Datadog without taking on full operational burden.

Adopt OpenTelemetry from day one

saves Avoids future migration cost

Instrument with OpenTelemetry rather than vendor-specific SDKs. Data flows to any backend that supports OTLP. Future platform switches drop from months to days. Future-proofs against vendor lock at the SDK layer.

Audit quarterly

saves Sustains all of the above

Cost growth that outpaces infra growth is the leading indicator of a problem. A quarterly cost review with a single owner catches new cardinality, new log volume, and unintentional retention upgrades before they become invoices.

Implementation

A seven-step roadmap

The order matters. Cut volume before you migrate platforms. Audit before you negotiate.

1
Audit current spend
Itemise spend by category. Identify the single largest line item.
2
Cut log volume first
Filter and sample at source. The fastest, lowest-risk saving.
3
Cap custom metric cardinality
List top 10 metrics. Remove high-cardinality labels.
4
Sample APM traces
Head-based 10 percent or tail-based on errors and slow paths.
5
Set up OpenTelemetry
Decouple instrumentation from vendor. Future migrations get cheaper.
6
Run Grafana Cloud or Prometheus in parallel
Validate parity for 30 days before any cutover.
7
Negotiate or migrate
Either renew with negotiated rates and a smaller floor, or cut over.

Quick win, this week

Audit your log volume by source. Add a drop rule for the noisiest non-actionable source. Most teams cut 10 to 20 percent of log spend in a single afternoon.

Quick win, this quarter

Run a custom-metric cardinality audit. Identify the labels generating the top three time series counts. Aggregate or drop. Typical impact: 20 to 30 percent on the metrics line.

Where to go next

Hidden costs →

What you are paying for that the pricing page does not list.

Run the calculator →

Model the saving from each strategy.

Open source TCO →

When migrating actually saves money, and when it does not.

Frequently asked

How can I reduce my Datadog bill?

Filter logs at source (30 to 50 percent saving), cap custom metric cardinality (20 to 40 percent), sample APM traces at 10 percent (15 to 30 percent), right-size retention (10 to 20 percent), and negotiate an annual commitment (15 to 25 percent off list). Stack-rank by current bill composition and target the largest line first.

How can I reduce monitoring costs by 50 percent?

Combine the four highest-impact strategies: log sampling, metric cardinality control, APM sampling, and retention tiering. Most teams report cumulative savings of 35 to 55 percent within a single quarter without changing vendors.

Should I switch to open source?

Only if you have platform engineering capacity. The licence saving (60 to 90 percent on metrics) is real but offset by infrastructure and engineering cost. The TCO crossover point is roughly 100 hosts with one engineer-quarter of setup. See the open-source-vs-paid TCO page.

Do annual contracts always save money?

List discounts of 15 to 25 percent are typical. The risk is committing to a usage floor that exceeds actual usage. Negotiate the floor down, secure true-up flexibility, and lock in exit terms before agreeing the discount.