Reality check: what people say they pay

What we found

• 3,478 documents (posts, comments and web pages) were found by searching for each of the 30 costed vendors except Chronosphere (now Palo Alto Networks Cortex XCOR), which is quote-only, so there is no list price to compare a bill with. Code narrowed 8,513 passages to 451 that state a US-dollar amount, are dated 2024 or later, and name a costed vendor.
• 21 reports passed the keep threshold for "reports an actual amount someone paid".
• 1 could be priced. Most reports give several amounts and no single bill, or no workload size such as hosts or GB of logs. Without a size there is nothing to price.
• Cost: $1.73 for searching and fetching through Glasser, and $0.24 for Jev.

Reported spend against our list estimate

Monthly USD, log scale. The circle is the bill the post reports. The square is this site's list-price estimate for a workload of the size it states. A dashed square is a floor: some of the vendor's prices are not published, so the true list total is at least this, and the reported ÷ list ratio is at most the one shown.

Splunk (Cisco)'s list estimate is a floor: it leaves out Custom metric series beyond 100 per host and Logs in Splunk Cloud Platform (GB/day committed), which Splunk (Cisco) does not publish a price for, while log volume is the size the post states, so the comparison mainly shows that the bill exceeds the part that could be priced.

So few points say nothing general about any vendor. A yearly figure is divided by 12. Where the post states only log volume, the rest of the workload (hosts, metrics, traces) is scaled from the site's reference workloads, so the estimate is approximate.

Reports we could not price

These passed the keep threshold and name a costed vendor, but code could not turn them into a priced point. We show no amounts for them: about one kept report in five is not a real bill (see the threshold below), and an unchecked figure next to a vendor's name could mislead.

Discount mentions

3 kept passages mention a negotiated discount, committed-use deal or custom contract (Datadog: 3). None of them states a discount rate we could use. Many people who have a discount don't mention it.

Method

1. Search. For each costed vendor we ran the same three queries ("vendor bill per month", "our vendor cost hosts", "vendor pricing expensive"). Searching and fetching went through Glasser, a pay-per-call data API service: Google web search and page fetches from Serper, and Reddit search and thread fetches from ScrapeCreators. Glasser has no Hacker News endpoint, so Hacker News posts came from Hacker News's own free public search API (run by Algolia), not through Glasser. No site was scraped directly. The same post found by several queries counts once.
2. Screen (code only). Code split each document into passages, then kept those with a currency amount (2,174), in US dollars only (2,069; other currencies were excluded, not converted), dated 2024 or later (891; undated pages were dropped), and naming a costed vendor (451).
3. Score. Jev (TypeSafe AI), which returns typed answers with probabilities and never numbers or text, answered five questions about each passage: does it report an amount someone actually paid (yes/no), which vendor (one of the list, or other), does it state a workload size (yes/no), is the amount monthly, yearly or unclear, and does it mention a negotiated discount (yes/no).
4. Threshold from a person. We ranked passages by Jev's probability of "actually paid". The site's owner labelled 27 passages spread across that range, and code chose the lowest probability at which at least 80% of the labelled passages at or above it were real bills. That is 0.51, with a precision of 82% (9 of 11).
5. Check and price (code only). A kept passage counts only if it names the vendor Jev chose, states a single dollar amount, and gives a period that agrees with Jev's answer. Code reads the amount and any stated sizes, builds the closest workload, and prices it with the same engine as the calculator.

What it cost

Spend was capped in code and every response was cached, so a rerun costs nothing. Most of the Jev calls come from a first pass that scored passages before the date and currency filters existed. The research calls Jev directly, not through this site's own spend limit.

Caveats

• Anecdotal. These are public posts people chose to write, not a sample. Nothing here shows what a typical customer pays.
• Possibly old. Posts from 2024 on are priced at today's list prices, and prices change.
• Negotiated prices. Large customers often pay less than list, and most don't say so.
• More than we price. A bill can include products, storage, add-ons or support the model does not cost, so a reported bill can exceed the estimate for that reason alone.
• Few usable points. People rarely state both a bill and a workload size in one place, which is what a comparison needs.
• Imperfect filtering. At the chosen threshold about one kept passage in five was not a real bill in the labelled sample, and that sample is small (27 passages).

What this does not change

Every number on this page and in the calculator comes from code. Jev only answered yes/no and pick-one questions, and never produced an amount. The site works fully without it. This page shows numbers, dates and links only: no usernames and no quoted text. It was generated on 2026-10-16. See the methodology for how list prices are computed.