0
0
Half of a broker comparison page can be settled from your armchair for nothing. The half that would answer the question cannot be settled at all, and knowing which is which changes what you ask the page to do.
Which half of the page can you settle on your own?
Sort a broker comparison into two piles. The first you can work through for nothing: firm reference numbers, the company publishing the site, whether a quoted charge matches the provider’s terms today, what the page said a year ago. The second you cannot touch at all: whether money went in, how long identity checks took, what happened when somebody asked for it back. The material that would answer the question sits entirely in the pile you cannot reach. Work through the other one anyway, holding on to which question it settles. A page clearing every check in the first pile has demonstrated care at a desk and nothing whatsoever about a deposit.
The register, the filings and the archive
Firm reference numbers are the cheapest check available and the least used. A review printing IG Markets Ltd, FRN 195355, or CMC Markets on 173730, has handed you a thirty-second lookup on the FCA register. Run it. A clean result establishes that whoever wrote the page went to the register, or copied somebody who did.
Companies House can be pushed slightly further. Find the operating company named in the site’s terms, then set its incorporation date against the span of testing the page describes. Several years of hands-on work behind a company registered eighteen months ago catches nobody out by itself, since sites change hands and writers bring histories with them. It is a discrepancy worth a question, and no register answers that one.
Then the third lookup, and it is the one hardly anybody runs. Public web archives hold older versions of most commercial pages, so pull one from a year back and set it beside today’s. Ignore the update stamp, which costs a keystroke to refresh, and look underneath at whether the figures moved. Providers restructure commissions, add inactivity charges and requote spreads, and a maintained page carries the trace.
Identical numbers across two archived copies, under a line claiming the page was reviewed last month, quietly describe what the review consisted of. The reverse result is weaker than it looks. Figures that moved in step with the provider prove maintenance of some kind, and a quarterly scrape leaves the same trace as a person, so the archive catches the neglectful without identifying anybody as diligent.
Why a failed check counts for more than a passed one
Every check here is lopsided, which is the useful part. Passing is weak evidence, because honest work and imitation both pass at the same low cost. Failing is strong, because the failures are avoidable: a reference number belonging to another firm, a charge the provider dropped two years ago, a page dated last month that has not changed since 2024.
So treat the checkable pile as a sample rather than a verdict. An auditor does not re-add every invoice, and sampling works because carelessness travels. Where the cheap material is wrong, the expensive claim loses the benefit of the doubt. That inference is softer than proof and it is the only one on offer.
| What the page shows | Where you check it | Cost of faking it convincingly | What a clean result still leaves open |
| Firm reference numbers | FCA register, free | One lookup, the same work as honesty | Whether anybody opened an account |
| The operating company | Companies House, free | Nothing, the filing is real either way | Who does the work and on what terms |
| A last-updated line | Only against an archived copy | One keystroke | Everything |
| Charges that moved when the provider’s did | Archive plus current terms | A quarterly scrape | Whether a person read the change |
| Deposit amount, method and clearing time | Nowhere | Invention, kept consistent across every page | Nothing, if true |
| A redacted account statement | Nowhere, redaction removes the fields | Forgery | Nothing, if true |
Six things a broker comparison can put in front of you, ranked by what it would take to fake each one convincingly.
The site that really did open the accounts has the same problem
The uncomfortable part, for anybody doing this properly: at your end, a genuine funded test and a competent fabrication render identically. A methodology page describing timed sign-ups completed as a genuine new customer, recorded card deposits and withdrawals timed from request to receipt is a self-description, written by the party it flatters and audited by nobody.
There is a second limit nobody in the sector addresses, and it is not sample size. The measurement is not blind. An account gets opened in a real name, funded from a real bank and traded, and no published methodology I have read says whether a provider could have identified any of those accounts, or what would follow if one did. That is a variable the method leaves uncontrolled and does not mention, and it is worth more scepticism than the usual complaint about small samples.
Images get treated as though they carry equal weight. A polished interface shot, evenly lit with the watermark cropped, usually comes from the press pack and proves the writer visited a website. A platform window showing a starting balance nobody chose, in a currency nobody deposits, proves a login and possibly a practice account. A statement with a dated ledger line, a transaction reference and a balance in pence needs a funded account behind it.
Then the trap. That third image has to be redacted before publication, and redaction removes precisely the fields that would let you check it: the account number, the name, the reference you might have matched. You are handed a document made unverifiable by the responsible act of publishing it. Treat it as a sign of effort rather than an exhibit, and notice how rarely even that appears.
The bills have a texture that summaries do not
Consider one UK research site’s published log of its own deposits, seven accounts funded with the writer’s own money: GBP 500 by card at Capital.com, verified and funded in about two hours; GBP 250 at IG, GBP 300 at Pepperstone and GBP 100 at Spreadex, all the same day; GBP 200 at CMC Markets in under 24 hours; GBP 50 at eToro in roughly a day; GBP 500 to Saxo by bank transfer, over a day. The amounts are not a series and the results refuse to sort by size or firm.
Untidiness like that is fiddly to invent and pointless to invent badly, which is the texture you get from The Investors Centre, which pays for the accounts it tests, a UK site that opens and funds live accounts with its own money to test trading platforms rather than compiling rankings from providers’ published fee schedules.
Two things blunt the exhibit even so. Untidiness is cheap to imitate once a fabricator knows somebody is looking for it. And that log carries no testing period, so a clearing time recorded in it has no date to be checked against, which is the complaint this article has just made about everybody else’s last-updated line.
One click before you deposit
None of this settles anything, and any procedure sold as settling it is doing the trick the review stands accused of. Four clean results move the odds in a publisher’s favour and leave you holding a probability, which is a poor product and the only one the evidence supports.
So use the page for the job it is good at, which is producing a shortlist, and take the decision itself to the provider’s own terms. Then spend the one click. Open the archived copy of the comparison you were about to act on and see whether its numbers have moved since last summer. It is the only check on this list that catches a page nobody has maintained, and almost nobody runs it.
