List-price estimates, not quotesNot affiliated with or endorsed by any vendor
Start from an illustrative workload, or describe your own. Every vendor below is priced for the same deployment: all logs and spans searchable for 30 days, every metric series stored. Prices come from each vendor's public rate card and are estimates, not quotes.
An illustrative reference workload, not a measured one.
Trace sampling
3,000 GB
logs a month
1,000M
spans a month
300,000
active series
100
hosts, 50 traced
= 800 GB of spans · 6,000M log events · 12,960M metric samples a month
Estimated monthly bill, cheapest first
Full estimateAt least: a needed price is unpublished
$10$100$1.0k$10k$100k
Bars use a log scale: each gridline is ten times the one before. Select a vendor for its line items, each price's source and date, and what was assumed.
Telemetry pipelines
Pipelines sit in front of a backend and cut what it bills for by filtering, sampling and routing. Their cost is added to a backend's, not compared with it.
In the dataset but not costed here (priced per trace, request or token, or not yet modelled): Arize AI, Helicone, LangChain, Langfuse.
vendors costed
30
list prices, each linked to its source
523
pages cited
762
dataset version
2026.10.07
How to read it
Illustrative workloads
The Small, Mid and Large presets are reference workloads, not measurements of any real system. Change any number to describe your own.
Like with like
Every vendor prices the same standard deployment, so a vendor whose customers usually sample or index less will look expensive here. What is priced.
Never zero by omission
A needed line item with no vendor-stated price makes the total a floor ("at least"), and the missing item is named. A vendor that publishes nothing priceable is shown as quote only, not as cheap.
Confidence
High: read off a published rate card; only the workload is assumed. Medium: rate card plus one or two stated conversions or allowances. Low: rests on our own sizing or conversion assumptions (compute sizing, bytes per sample).
Not included
RUM, synthetics, LLM observability, security and SIEM, profiling, on-call, support plans, egress and taxes. Negotiated prices are usually lower than list.
Open data
Every price, source and formula behind these totals is published as JSON and CSV with a schema. Get the data.