The standard deployment
Every vendor is priced for the same deployment, so totals compare like with like:
- infrastructure monitoring on every host;
- tracing on the hosts that run traced services, with every span kept and searchable for 30 days;
- every log ingested and searchable for 30 days;
- every active metric series stored at the vendor's standard metrics retention. OpenTelemetry and Prometheus metrics count as custom metrics wherever a vendor makes the distinction;
- the stated number of people with access to dashboards and queries.
Many customers sample traces or index only part of their logs, which costs less. A vendor whose customers usually do that will look expensive here. That is deliberate: a sampled deployment is a different, cheaper workload. Use the sampling control to see it.
The annual-commitment list price is used wherever one is published, and the on-demand price otherwise.
What is not included
Real-user monitoring, synthetic checks, LLM observability, security and SIEM, continuous profiling, on-call and incident tools, support plans, network egress and taxes. Negotiated discounts are not modelled, and they are usually substantial.
Where prices come from
- A price appears as a number only when the vendor's own pricing page or documentation states it. Each price links to that source and shows the date it was read.
- A price that only a reseller, a marketplace listing or another third party states is shown as unpublished.
- An unpublished price is never counted as zero. A total that needs one is a floor, shown as "at least" with a dashed outline instead of a solid bar, and the missing item is named.
- A vendor that publishes nothing priceable for this workload is shown as quote only, not as cheap.
- No vendor text is reproduced beyond product and price-item names and the titles of the pages cited.
How a bill is computed
Each vendor's costing lists the line items a standard deployment uses. A formula turns the workload into a monthly quantity in each item's unit: GB of logs, millions of spans, hosts, series, seats. Free allowances are subtracted, volume tiers price only their own band, and a platform minimum replaces a smaller usage total. Derived quantities use fixed conversions: span volume is spans × average span size, log events are log volume ÷ average event size, and metric samples are series × samples per month at the scrape interval. GiB-priced items convert at 1.073741824 GB per GiB, and an hourly price uses 730 hours a month.
Confidence says how much of a total rests on assumptions rather than the rate card: high means read off a published rate card; medium means a rate card plus one or two stated conversions or allowances; low means it rests on sizing or conversion assumptions such as compute sizing or bytes per sample.
Sampling. Sampling in your own collector drops spans before they reach the vendor, so every span charge shrinks. When the vendor samples, charges it meters on everything it receives still apply to every span. Those charges are marked in the breakdown, and vendors whose cost for dropped spans is not published are left out of that view.
The workloads
Small, Mid and Large are illustrative reference workloads, not measurements of any real system. Every number can be changed in the calculator, and the page's address records the change so a link reproduces the view.
Reporting a wrong price
Each vendor's breakdown has a "This price is wrong" link. It opens an issue in the public repository, prefilled with the vendor, the line items and the current view. Please include a link to the vendor's own page that states the correct price. Vendors are welcome to report corrections in the same way.
The data
Everything the calculator uses is published as JSON and CSV with a JSON Schema for each: the data page lists the files.
Privacy
This site uses no analytics, no cookies and no third-party requests, and stores nothing in your browser. The view you build lives only in the page's address.