How firms are assessed

Principles

• We assess what can be verified

  Assessments rest on public records, the firm’s own documents and checks we can repeat. We do not rely on rumour or reputation.

• Every judgement cites evidence

  An answer of met, partly met or not met counts only if it cites a source. Without one, the criterion is treated as not determined.

• Unverified is not the same as met

  Below the minimum evidence threshold we still score the firm, but anything we could not verify earns no credit, and the score is marked as a limited-evidence score. “Not yet assessed” is never shown as 0/100.

• Missing evidence is not a finding

  In a full score, what we could not establish is left out and lowers our confidence instead. Either way it is never reported as a finding against the firm. Absence of a disclosure we checked for is different, and counts as not met.

• Scores are calculated, never typed

  The number is computed from the stored answers and findings by the rules on this page. Nobody, including us, can enter or adjust it.

• Versions never rewrite history

  Each assessment is pinned to the methodology version it used and stored with its inputs. A later change to weights affects only new assessments.

How the score is calculated

The score runs from 0 to 100. It is a weighted average of nine dimensions. Six are scored from criteria; the enforcement dimension is scored by deductions for substantiated adverse findings; and two come from our on-site assessment.

Criteria

Each criterion carries points within its dimension. We answer it, citing evidence:

A dimension is scored only when at least 50% of its applicable points are answered with evidence. Its score is the share of answered points earned. In a full score, dimensions that are not scored drop out and the remaining weights are rescaled. Below the evidence threshold the rules for a limited-evidence score apply instead (see the threshold section). The overall score is rounded to the nearest whole number after weighting.

The nine dimensions

Enforcement deductions

The enforcement dimension starts at 100. Each substantiated adverse finding deducts points by severity, scaled down as it ages, and the dimension cannot fall below zero. A finding counts only when it is one of the types and statuses below and at least one of its sources is a regulator, court or official record.

What counts:

• Regulator warning: active
• Enforcement action: concluded or settled
• Legal finding: final

Allegations, complaint records and clone warnings never count. Neither do matters that are ongoing, under appeal, overturned, dismissed or removed. They are still shown on the profile, with their status, so you can weigh them yourself.

Finding types and labels

We label each finding by what it is. These labels are never merged into a single verdict.

Evidence standard

Every source is recorded with its title, publisher or issuing authority, address, publication date where known, the date we checked it, the jurisdiction and legal entity it concerns, and a note on what it supports. Where possible we also keep an archived copy.

• A firm’s own documents are good evidence of what it promises clients, and weak evidence of anything else.
• Information from members of the public is a lead. It is kept private and never supports a published finding on its own.
• We do not publish a regulator warning, enforcement action or legal finding unless we have checked it against the issuing body’s own record.

On-site verification points

Most assessments are research-based. A company can also complete the TTI On-Site Verification & 30-Day Platform Review: an office visit, meetings with company staff, identity and regulatory-record checks, and structured testing of a specified platform over at least 30 days. How the programme works.

• Research-based assessment

  An assessment using documented public information and available sources. This does not imply an office visit or hands-on platform testing.

• TTI On-Site Verification & 30-Day Platform Review

  An assessment involving a physical office visit, meetings with company staff, identity and regulatory-record checks, and structured product testing over 30 days.

How the points are earned

• 15 of the 100 points are for what we verify in person: 5 for what our visit confirms about the office of the assessed legal entity, and 10 for the outcomes we record across the ten categories of our on-site checks and 30-day testing, each scored met, partly met or not met.
• The office checks, recorded with the visit:
  • Real office at the stated address, not a ghost address: 2 points
  • Office of the assessed entity: 2 points
  • Staffed and operating: 1 point
  A ghost address, such as a mail drop, a virtual office or an empty unit, earns nothing for that check and blocks the badge.
• The office points count once our visit report is published. The testing points count once the review’s decision is recorded, after the private findings discussion in which the firm can point out anything we missed.
• A firm we have not checked in person scores zero for the office visit. Until we have completed an on-site assessment, the testing points are not scored and their weight is shared among the other dimensions, so no firm is marked down for testing that has not taken place.
• On-site assessments are carried out through our verification programme. The company pays for the work, and every report says who paid; the fee earns no points. The points follow only what we find, so a check we cannot confirm earns nothing and a poor result scores low.
• What we observe on site can also be evidence for the research criteria, for example whether the entity named in the account agreement matches the register. Where it bears on a criterion, the criterion’s note says so.
• The badge decision is separate from the score. It follows the programme’s published requirements, so a firm can hold the badge with a modest score, or be refused it with a high one.

Testing points

Each point is reported as “tested or checked”, “partially tested”, “not tested” or “not applicable”. Every point that is partially tested or not tested carries a short explanation.

Minimum evidence threshold

A full score requires all of these. The threshold and every reason it was or was not met are published with each assessment.

• These dimensions are assessed: Regulatory status and licence verification and Enforcement and substantiated adverse findings.
• Assessed dimensions carry at least 60% of the total weight , leaving out on-site work we have not carried out.
• At least 2 independent primary sources (regulator, court or official records) are cited.
• At least 1 of them is a regulator record.

Limited-evidence scores

When the evidence does not meet the threshold, we still publish a score, marked as a limited-evidence score, with the threshold checks that were not met. In it:

• every dimension counts, at its full weight;
• criteria we could not verify earn no credit, and a dimension we could not assess at all scores zero;
• on-site testing that has not taken place is still left out, as in a full score.

A firm that publishes clear, checkable information is easier to verify and can score higher. A limited-evidence score says that we could not confirm enough of the record; it is not a finding that the firm did anything wrong. If we could verify nothing at all, the assessment is published as “Not scored”.

Confidence level

Confidence tells you how much weight the score can bear. It is calculated separately from the score and does not change it. It multiplies two things:

• Completeness: the share of the methodology, by weight, that we could answer with evidence , leaving out on-site work we have not carried out.
• Evidence strength: the quality of the best source behind each answer (60%) and how recently it was checked (40%). Evidence checked within 120 days of the review counts in full, falling to nothing at 730 days.

Because it multiplies, confidence can never be higher than the share of the methodology we assessed.

Score bands

Each score carries a word, so the number is never read alone. The words describe the evidence, not the future.

Updates and stale evidence

• Every assessment shows its review date. After 365 days it is flagged as due for review.
• A cited source last checked more than 365 days ago is flagged as stale on the profile.
• We re-review early when something material happens: a new warning or enforcement action, a change of licence or ownership, credible new evidence, or a firm’s response that points to evidence we had not seen.
• A new assessment supersedes the old one. Superseded assessments stay available, exactly as published, from the profile’s history.
• If an assessment proves wrong, we withdraw it, say why and log it in our corrections.

What we do not assess

• Whether a firm is a good choice for you, or whether trading suits you at all.
• Prices, spreads, platforms, apps or customer service quality, except where they are matters of disclosure or part of the on-site testing points. For a firm we have not assessed on site, the score does not rate them.
• Investment performance, returns or the merits of any product.
• A firm’s solvency or the likelihood that it will fail. We are not a credit rating agency.
• Activities outside the firm’s trading services, and firms outside the categories we cover.

Limitations

• We assess what is public or provided to us. We cannot audit internal controls or see how client money is actually held day to day.
• Registers and warning lists can lag behind events, and regulators publish in different ways. We record what each showed on the date we checked.
• A firm with no recorded findings may still have problems we have not found. “No findings recorded” never means “no problems”.
• Weights reflect our editorial judgement of what matters most. We publish them so you can disagree, and we review them openly.
• Scores under different methodology versions are not directly comparable.

Versions and changes

The methodology is versioned. A published version is never edited: changes create a new version, which applies to new assessments only. Each assessment states the version it used.

1. 7 Oct 2026

   Version 1.2 (current). First version in force. Seven research dimensions weighted 26, 13, 13, 13, 8, 8 and 4, plus on-site verification: 5 points for what our office visit confirms, counted once the visit report is published, and 10 for the outcomes of on-site and platform testing, counted once the review is decided. A partly met criterion earns 75% of its points, and substantiated enforcement findings deduct 2, 10, 25 or 50 points by severity before ageing. A full score needs 60% of the weight assessed; below that, a firm receives a limited-evidence score, in which anything we could not verify earns no credit. A separate confidence level is based on completeness, source quality and freshness. Before launch we recalibrated the unpublished draft 1.1 for trading firms, which are not held to the standards of a bank: partial credit rose from 50%, deductions fell from 5, 15, 35 and 60 points, and the evidence threshold fell from 70%. Criteria, weights, score bands and the size of the on-site points are unchanged; the office points no longer wait for the platform test.

Independence rules are part of our editorial standards. To suggest a change to the methodology, contact us.