Why does a monitoring bill arrive looking nothing like the quote that sold you the tool? It happens constantly, and the reason is rarely a billing error. Application monitoring tools price in fundamentally different ways, so the number you compare at the start is often not the number you pay at the end.
This comparison breaks down how the major platforms actually charge, where budgets quietly leak, and how to model a real workload before signing anything.
Why Headline Prices Mislead Buyers
Most buyers compare a single number: cost per host, cost per gigabyte, cost per seat. That is like comparing the price of a car by looking only at the tires. Each vendor picks a different unit to meter, and the unit determines everything about your final bill.
There are four main pricing structures in this market:
| Pricing Model | How It Works | Typical Vendors |
|---|---|---|
| Per host | Flat monthly fee for each server or VM monitored | Datadog (infrastructure) |
| Per GB ingested | Charge for the volume of logs, traces, and metrics sent | New Relic, Grafana Cloud |
| Consumption units | Bundled “host units” or similar credits consumed by features | Dynatrace |
| Per seat plus usage | A fee per user, layered on top of data charges | New Relic and several mid-market tools |
A workload with 40 hosts, heavy log volume, and 12 engineers will produce wildly different totals under each model. The only fair comparison is to price your own environment against all four.
Datadog: Per-Host Pricing With a Metric Trap

Datadog built its reputation on infrastructure monitoring priced per host. In recent years, list pricing has sat at roughly $15 per host per month for Infrastructure Pro and about $23 per host per month for Infrastructure Pro Plus on annual billing. Application performance monitoring (APM) is sold separately, at roughly $31 per host per month for APM and about $40 for APM Pro.
That separation matters. A team that wants both infrastructure and APM visibility is looking at roughly $46 per host per month before anything else. For 50 hosts, that is about $2,300 monthly, or close to $27,600 a year, on list prices alone.
Then comes the part that catches people: custom metrics. Datadog includes an allotment of custom metrics per host, and charges beyond that, historically around $0.05 per 100 custom metrics per host per month. That sounds trivial until a team instruments a high-cardinality dimension, like user ID or request path, and suddenly generates hundreds of thousands of unique metric series.
Practical tip: audit metric cardinality before you migrate, not after. Pull your existing metric list, count unique label combinations, and multiply. Teams routinely discover that two or three careless tags account for most of their overage.
New Relic: Usage-Based Pricing Rewards Discipline
New Relic changed the conversation in 2020 when it moved to a usage-based model. Instead of charging per host, it charges per gigabyte of data ingested, plus a fee per full user. Historically that has meant roughly $0.25 to $0.35 per GB depending on tier, and around $99 per full user per month, with a free tier of 100 GB per month and one full user.
The appeal is flexibility. Ten hosts sending little data cost less than one host sending a flood. Small teams can often stay inside the free tier indefinitely, which makes New Relic a common entry point.
The risk is the opposite of Datadog’s. When data volume is the meter, verbose logging becomes expensive. A debug-level log stream left running in production can move a bill by thousands of dollars in a month. The discipline required is real: sampling, log level management, and retention rules are not optional extras but core cost controls.
A practical habit that works: set a data budget per service, not per team. When one service owns its own ingestion allowance, the engineer who adds the noisy log line feels the consequence directly.
Dynatrace: Consumption Units and Enterprise Weight
Dynatrace prices through consumption units, historically built around a “host unit” concept. Full-stack monitoring has commonly been cited at roughly $69 to $80 per host unit per month, with infrastructure-only monitoring priced lower.
On paper, that looks like the most expensive option, and for simple infrastructure monitoring it often is. But Dynatrace bundles considerably more into its full-stack rate, including automatic discovery, AI-assisted root cause analysis, and distributed tracing without separate line items. The comparison only becomes fair when you total what a Datadog or New Relic deployment costs after adding the equivalent modules.
The model has also shifted toward more granular consumption metrics over time, so verify the current structure rather than relying on older host-unit math. Enterprise buyers should also expect annual commitments and negotiated rates, which can move the effective price well below list.
Open Source: No License Fee, Real Costs Elsewhere
Prometheus for metrics, Grafana for dashboards, and Jaeger or Tempo for tracing carry no license fee at all. Prometheus is a graduated project of the Cloud Native Computing Foundation, having reached that status in 2018, and it is the default metrics layer in most Kubernetes environments.
The catch is that “free” describes the software, not the operation. Self-hosting shifts cost into three buckets: infrastructure for storage and compute, engineering time to run and upgrade the stack, and the opportunity cost of what those engineers are not building instead.
A rough rule that holds up in practice: below roughly 20 to 30 hosts, open source is usually the cheapest path if you already have platform engineering capacity. Above that, the operational burden grows faster than host count, and a managed service often wins on total cost even at a higher sticker price. The break-even point depends almost entirely on how expensive your engineers are and how much they enjoy running storage clusters.
Grafana Cloud: The Managed Middle Ground
Grafana Cloud sits between self-hosted open source and full-suite commercial APM. It offers a free tier and usage-based paid tiers that bill metrics, logs, and traces by volume or series count.
For teams already comfortable with the Grafana and Prometheus ecosystem, this is often the lowest-friction managed option. You keep the query language and dashboards you know, and someone else handles storage and uptime. The trade-off is that Grafana Cloud is a monitoring and observability platform rather than a full APM suite with the deep code-level profiling that Datadog, New Relic, and Dynatrace provide. Teams that need tight application performance tracing may end up pairing it with another tool, which reintroduces the multi-vendor cost problem.
The Hidden Line Items Nobody Quotes
Headline rates rarely include the costs that decide the final bill. Watch these closely:
Retention and egress. Storing 30 days of logs costs far more than storing 7. Vendors price retention tiers aggressively, and pulling data out for analysis can carry its own charge.
Custom metric cardinality. As noted with Datadog, unique label combinations multiply fast. A single high-cardinality tag can double a metrics bill.
Log volume. The single most volatile line item in usage-based pricing. One misconfigured service can out-log the entire rest of the estate.
Seat counts. Per-user fees add up quietly. A 30-person engineering organization on a $99 per-user plan is paying nearly $36,000 a year before a single gigabyte of data.
Overage rates. The per-unit price after you exceed your committed volume is usually much higher than the committed rate. Ask for it in writing.
Contract terms. Annual commitments usually earn a discount but lock in volume estimates. Under-forecast and you pay overage; over-forecast and you pay for capacity you never used.
How to Model a Real Comparison
List prices cannot answer “which is cheaper for us.” Only your own workload can. Follow these steps before requesting quotes:
- Count hosts, containers, and serverless functions you actually need to monitor.
- Measure daily log ingestion in gigabytes over a typical week, not a quiet one.
- Count unique custom metric series, including every label combination.
- List how many engineers need full, write-level access versus read-only dashboards.
- Decide retention requirements per data type: metrics, logs, and traces rarely need the same window.
- Price all four models against those numbers, then add 20 percent headroom for growth.
That last step matters more than it sounds. Observability spending tends to grow with success: more services, more traffic, more data. A plan that fits exactly today will be over budget in six months.
Common Mistakes That Inflate Bills
Three errors show up again and again in real deployments.
The first is instrumenting everything by default. Teams enable full tracing and debug logging across all services because it feels thorough, then discover that 80 percent of the data is never queried. Sampling and tiered retention fix this cheaply.
The second is ignoring cardinality during design review. A tag like request ID or customer email should never become a metric label. Catching that in code review costs nothing; catching it on the invoice costs a quarterly budget conversation.
The third is comparing quotes instead of total cost of ownership. A vendor that looks 30 percent cheaper per host may charge separately for dashboards, alerting, or synthetic checks that a competitor bundles. Build a line-item spreadsheet covering every capability you need, then compare totals.
Which Tool Fits Which Team
A few patterns hold up across most organizations.
Small teams under 20 hosts with limited budget usually do best starting on a free tier, either New Relic’s 100 GB allowance or Grafana Cloud’s free plan, and graduating only when limits bite.
Mid-sized engineering organizations with predictable infrastructure often find per-host pricing easier to forecast, which favors Datadog despite the metric overage risk, provided cardinality is controlled.
Teams with highly variable or spiky workloads tend to prefer usage-based pricing, since they pay for what they actually send rather than peak capacity.
Enterprises needing deep automatic discovery and AI-assisted analysis across complex estates often accept Dynatrace’s higher list price in exchange for bundled capability and fewer integration projects.
Platform teams with strong Kubernetes skills and existing engineering capacity frequently run Prometheus and Grafana themselves and treat the saved license fee as headcount for the platform team.
Conclusion
The honest answer to “which application monitoring tool is cheapest” is that the question cannot be answered without your own numbers. Per-host pricing favors stable, well-understood infrastructure. Per-gigabyte pricing favors disciplined logging. Consumption models bundle more but charge accordingly. Open source removes the license fee and adds an engineering bill.
The teams that win this comparison are not the ones who negotiate hardest on the headline rate. They are the ones who measure their own workload first, model all four pricing structures against it, and build in headroom for growth. Do that, and the invoice that arrives twelve months later will look like the one you planned for.
Frequently Asked Questions
What is the cheapest application monitoring tool for small teams?
For teams under roughly 20 hosts, free tiers usually win. New Relic’s free allowance of 100 GB per month with one full user, and Grafana Cloud’s free plan, both cover small production environments at zero cost. Self-hosted Prometheus and Grafana are also free on license but require engineering time to operate.
Why is my Datadog bill higher than the per-host quote?
Almost always custom metrics. Datadog includes an allotment of custom metrics per host and charges beyond it, historically around $0.05 per 100 custom metrics per host per month. High-cardinality tags, such as user IDs or request paths used as labels, can generate hundreds of thousands of unique series and dominate the bill.
Is usage-based pricing cheaper than per-host pricing?
It depends entirely on your data volume. Usage-based pricing rewards lean, well-managed telemetry and punishes verbose logging. Per-host pricing is easier to forecast but charges you the same whether a host sends one gigabyte or one terabyte. Teams with spiky or low-volume workloads usually benefit from usage-based models; stable, high-volume estates often prefer per-host.
What does Dynatrace cost compared with Datadog?
Dynatrace full-stack monitoring has historically been cited at roughly $69 to $80 per host unit per month, against Datadog’s roughly $46 per host per month for infrastructure plus APM. The gap narrows once you add the modules Dynatrace bundles by default, such as automatic discovery and AI-assisted root cause analysis. Always compare complete capability sets, not base rates.
Do open-source monitoring tools really save money?
They save license fees, not total cost. Prometheus and Grafana carry no license charge, but self-hosting adds storage infrastructure, upgrade work, and on-call responsibility. Below roughly 20 to 30 hosts with existing platform capacity, open source is often genuinely cheaper. Above that, operational overhead frequently erases the savings.
What hidden costs should I check before signing a monitoring contract?
Check retention tiers per data type, egress charges for exporting data, custom metric cardinality limits, per-seat fees, and the overage rate charged after you exceed committed volume. Also confirm whether alerting, dashboards, and synthetic checks are included or billed separately, and whether annual commitments lock in volume estimates you may outgrow.