FirmFax earns affiliate commission from some of the firms named in this study. That never changes what the data says. How we make money.

Price provenance · 2026-09-05

A quarter of the prices we list could not be found on the pages we cite

An audit of FirmFax's own data. We matched all 294 prices we list against the captured text of the pages we cite for each one; 77 did not appear. A hand check of 18 of those in a live browser put the method's false-positive rate at 11%. Both datasets are published as files.

This is an audit of FirmFax's own data. Method, limitations and both raw datasets are at the foot of the page.

77
of 294 prices we list
Not found on the page we cite for them
31
of those
Reconstruct as a published list price times a promo rate
35
of 71 firms
Have at least one price we could not source
11%
of a 18-price hand check
Were the method being wrong, not the price being absent

A prop firm evaluation is bought on price. It is the first number a trader compares and the one every comparison site, including this one, puts in its cheapest column. So we pointed the check at ourselves: for every price FirmFax lists, does that exact figure appear in the text of the page we cite as the source for it?

For 26% of them it does not. 77 of the 294 prices we store, spread across 35 of the 71 firms we track, are not on the page our own record points to.

That is a finding about our data before it is a finding about anyone else's. A price we cannot trace to the page we cite is a price a reader has no way to check, whatever the reason — the firm moved it, the firm never printed it, or we wrote it down wrong. We are not claiming these firms hide their prices. We are saying we could not source ours.

The arithmetic tell

31 of the 77 are not random. They reconstruct exactly as a list price that is on the page multiplied by a promotional rate — the signature of a number that was calculated rather than read.

The clearest case is AudaCity Capital. It runs a public 25% promotion and prints $37 $49 on its own homepage. FirmFax listed $36.75, which is 49 × 0.75 to the cent. Nobody pays $36.75. That figure was ours, not the firm's: the firm's own sale price is a quarter higher, and the number we published was one no trader could have been charged.

A discounted price is not itself the problem. Blue Guardian publishes "Account 5K $54.00 $72.00 … Save 25% with code BG25". The $54 is real, obtainable, and printed. The test is not whether a discount was applied. It is whether the number came from the firm or from a calculator of ours.

What we checked by hand

Matching text against a captured page is a blunt instrument, so we took a random sample of 18 of the 77 findings and opened each one in a real browser — first the page we cite, then the firm's own pricing page. 2 of the 18 were the method being wrong: the figure is printed on a page we cite and the check did not see it. That is a false-positive rate of 11%.

Applied to the whole run, that puts the number of genuinely unsourceable prices at roughly 68 rather than 77, or about 23% of what we list. A sample of 18 is small, so treat that as an estimate with real width, not a second decimal place.

A further 3 of the 18 (17%) were prices the firm does publish, just not on any page we cite — HyroTrader's $89 is on its homepage, which our record for that plan never pointed at. Those are still defects in our sourcing, and they are the reason this study can only speak about the pages we cite. It cannot tell you what is or is not on a firm's whole site.

The sample also turned up two errors the automated check could not name. Lux Trading Firm publishes £299, and we stored it as $299. FTMO charges in euros, and we stored a dollar figure. Both were counted as unsourced for the right reason and the wrong one.

Download the spot-check, all 18 rows (JSON)

Limitations

The check reads captured page text. Four things are invisible to it, and all four produce a false "not found":

  • Prices drawn by JavaScript. Not hypothetical: on 5 September a browser pass found Fintokei's payout figures render only once the page is scrolled, and read as $0 to a plain fetch. Dominion's certificates render only in a real browser. Text captured without running the page misses both.
  • Prices behind a checkout or a login. Several firms show a price only in the cart, after an account is created, or after a region is chosen.
  • Prices set in an image — a pricing table shipped as a graphic has no text to match.
  • Pages we never captured, including a cited URL that has since moved, 404ed, or been geo-redirected.

This is why the headline says "the pages we cite" and not "the firm's site". The 11% measured above is the size of the effect in one sample; it is not zero and we have no reason to think it is stable across firms.

Every price we could not source

All 77 findings from the run of 2026-09-05, as printed by the check. "Promo arithmetic" means the stored figure equals a list price on the page times a round discount. Rows in this table are claims about our data, and 11% of a sample of them turned out to be wrong in our favour.

Firm Plan We listed Why it failed

Download the full run (JSON)

What our records hold on rules

Rules fare better than prices, with the same caveat: these count what we have found and stored, so they are a floor. A firm missing from a count below may simply be one whose page we have not read closely enough yet.

We have a news blackout recorded for 33 of the 71 firms, and where the window is published it clusters tightly:

Blackout either side of a release Firms
1 minute 3
2 minutes 9
3 minutes 1
4 minutes 1
5 minutes 13
10 minutes 1

Consistency rules are similar. We hold a published percentage for 43 firms, median 40%. Rarer is the opposite claim: we have found only 4 firms stating plainly that no consistency rule applies. Every other firm without a percentage in our record is silent as far as we have read, which is not the same as having no rule — and treating it as the same is how a "no consistency rule" list ends up recommending firms whose own help pages document one.

A number that measures us, not the industry

Traders search for a firm's name followed by "payout proof" more than for most of its rules. We hold dated evidence — a ledger with amounts and dates, or an itemised case study — for 21 of the 71 firms.

That figure is a measure of our coverage and nothing else. It was 2 on the morning of 2026-09-05 and 21 by the evening, and no firm published anything new in between; we went looking. Anyone quoting it as "only N prop firms prove they pay" would be quoting our to-do list. The same warning applies to the counts above and to the 17 firms for which we have recorded a sourced concern and the 1 whose record we have checked claim by claim. They move when we work.

Method

Every price FirmFax stores was matched against the captured text of the pages our record cites for that price, allowing for how prices are written in practice ($49, 49.00, 49). Only the firm's own domain counts; offer marketplaces and coupon aggregators were excluded, because a third party's listing is not the firm publishing a price. A price of zero is skipped: it marks a plan earned by passing an evaluation rather than bought. The check is npm run check-prices in our repository, and the figures above are frozen from the run on 2026-09-05 — the same file linked below, so the page and the download cannot disagree.

The spot-check was drawn at random from that run's 77 findings and worked by hand in a browser, one firm at a time, recording separately whether the figure appears on a page we cite and whether it appears anywhere on the firm's site. Rule counts are computed from the live dataset at build time and will move as we add records.