Observability has a value problem before it has a price problem
Teams collect more telemetry than they can use, buy overlapping tools, and discover the economics only after traffic, cardinality, and retention rise. Six findings from the guide:
Cost is now a top buying criterion
In Grafana Labs' 2025 survey of 1,255 respondents, 74% named cost a top tool-selection priority; 39% called complexity and overhead their largest obstacle.
Most telemetry is never used
A 2025 Sawmills survey found only 13% of collected telemetry is actively used on average, and 84% of companies use less than a quarter of it.
Tools overlap everywhere
The average organization runs eight observability technologies. The direct cost is duplicated ingest; the larger cost is split context and duplicated work.
Same logs, wildly different bills
300 GB of monthly logs can land anywhere from ~$27 to ~$540 depending on whether the vendor meters GB ingested, events indexed, or storage retained.
One label can cost thousands
A single unbounded label (20 services × 10 regions × 5 codes × 50 tiers) creates 500,000 active series - roughly $3,185/month at Grafana Cloud's first public tier.
Low-value ingest is cuttable
The guide's 90-day program targets a measured 20–40% reduction in low-value ingest - as an operating goal with baselines and budget alerts, not a promise.
Nine platforms, seven different ways to meter a bill
Host-based, ingest-based, event-based, active-series, sample-based, stored-GB, query-compute, and hybrid plans. A condensed view - the guide covers what each buyer must still resolve.
| Platform | Public pricing signal (list, USD) |
|---|---|
| Datadog | Infra Pro from $15/host/mo; APM from $31/host/mo; log ingest from $0.10/GB plus $1.70 per million indexed events (15-day). |
| Splunk Observability | Infra from $15/host/mo; End-to-End $75/host/mo, billed annually; logs sold separately. |
| New Relic | 100 GB/mo free, then $0.40/GB (original) or $0.60/GB (Data Plus), plus users or compute. |
| Grafana Cloud | $19/mo platform fee; metrics $6.50 per 1,000 active series; traces $0.05/GB processing + $0.40/GB writing after 50 GB. |
| Honeycomb | Free to 20M events/mo; Pro from $150/mo for 50M events; collection from $0.10/GB; unlimited seats and queries. |
| Elastic Serverless | Log ingest as low as $0.07/GB; retention as low as $0.017/GB-month; egress beyond 50 GB at $0.05/GB. |
| AWS CloudWatch | No minimum; OpenTelemetry metrics $0.50 per uncompressed GB ingested with 15 months of storage; varies by region. |
| Coralogix | Frequent-search logs $1.15/GB; monitoring logs $0.50/GB; one unit = $1.50 of signal usage. |
| KloudMate | Free tier included; Pro $99/mo telemetry budget; logs and traces $0.40/GB, metrics $0.30/million samples, no host or user fee. |
Public list prices as of 17 September 2026. This is a pricing-model map, not an apples-to-apples ranking - the guide shows the metering boundary behind each number.
"Elastic at $0.09/GB vs New Relic at $0.40/GB" is not a 4.4x conclusion
One meters retention separately, the other bundles product access. Datadog mixes GB ingestion with event indexing; Grafana splits processing and writing; Honeycomb charges around events. The guide's fix: replay the same telemetry through both platforms for 30 days and reconcile both invoices.
Ten sections, from pricing map to 90-day program
Written for the people who own the bill
Founders & engineering leaders
You own an observability bill that grew faster than expected and want to know whether it's normal.
SRE & platform teams
You've been asked to cut telemetry spend without losing production answers.
FinOps practitioners
You're extending cloud cost discipline - allocation, forecasting, unit economics - to observability.
Anyone evaluating vendors
You want a comparison method that survives procurement scrutiny, not another "we're cheapest" claim.
From unexplainable bill to continuous control
Make the bill explainable
- Name a cost owner per service
- Build a daily unit-cost dashboard
- Attribute 80%+ of telemetry
- Set budget alerts at 50/75/90%
Reduce low-value volume
- Remove debug and duplicate logs
- Cap metric label cardinality
- Sample traces, always keep errors
- Split hot, archive, and compliance tiers
Make control continuous
- Telemetry budgets in deploy checks
- Detect volume regressions weekly
- Showback cost to teams
- Reprice from measured workload
Vendor-authored desk research, labeled as such
Read this before you cite us
This guide is built from public vendor pricing, vendor documentation, and published third-party surveys - not original fieldwork. Where a number comes from a vendor that sells cost optimization, the guide says so. Where platforms are compared, the metering boundary is shown instead of pretending the units match.
KloudMate sells an observability platform, and our own pricing appears in the benchmark table under the same normalization rules as everyone else. Treat our inclusion the way you would treat any vendor-authored comparison: a starting point, not a verdict. If our arithmetic is wrong, tell us - we publish corrections.
Prices were checked against vendor pages on 17 September 2026 and will drift. Re-verify before any purchasing decision.
Common questions
How much does observability cost per GB in 2026?
Public list rates range from about $0.07/GB (Elastic Logs Essentials ingest, volume-dependent) to $0.40/GB (New Relic original data; KloudMate logs and traces) and $1.15/GB (Coralogix frequent-search logs). Per-GB rates are not directly comparable - vendors meter different boundaries: uncompressed ingress, stored data, events, hosts, or active series.
Why do observability bills grow faster than expected?
Log verbosity multiplied by traffic, metric cardinality multiplied by scrape frequency, trace span fan-out, long hot-retention defaults, and overlapping tools. Grafana Labs' 2025 survey found teams run eight observability technologies on average, and 39% called complexity and overhead their largest obstacle.
How can we cut observability costs without losing visibility?
Make the bill explainable first (ownership, attribution, unit-cost dashboards), then remove low-value volume (debug logs, unbounded labels, unsampled traces), then make control continuous with telemetry budgets, showback, and measured repricing at contract anniversaries. The guide's 90-day program details each step.
Does OpenTelemetry reduce observability costs?
Not by itself. OpenTelemetry improves portability and routing control, but sending the same high-volume, high-cardinality data through a Collector merely makes expensive data portable. Cost reduction comes from deciding what to collect, sample, and retain - the Collector is where you enforce it.
Is this guide independent research?
No. It is vendor-authored desk research published by KloudMate, built from public pricing and labeled third-party surveys. KloudMate's own pricing is benchmarked under the same rules, and the conflict is stated plainly in Section 7 of the guide.
Every number is cited
All 15 sources are listed in the guide with access dates. The primary ones:
- Grafana Labs, Observability Survey 2025 (1,255 respondents)
- Sawmills, Telemetry Pipeline Survey 2025 (vendor data - directional)
- Datadog public pricing pages (Infrastructure, APM, Logs)
- Splunk Observability Cloud public pricing
- New Relic public pricing and data options
- Grafana Cloud public pricing (metrics, logs, traces, Adaptive products)
- Honeycomb public pricing
- Elastic Observability Serverless pricing
- AWS CloudWatch public pricing
- Coralogix public pricing
- KloudMate public pricing
- OpenTelemetry sampling documentation
- FinOps Foundation, State of FinOps 2025
Benchmark your bill before your next renewal
The full pricing map, workload arithmetic, procurement benchmark, and 90-day control program - free, with every source cited.
Download the guide →