How case files are assessed
Each case file is taken from a transcript of the podcast, written up as a case file and checked against published and trusted sources. The file checks facts and grades evidence.
A file assesses the strength of the evidence, not the popularity of the claim.
How a case file is produced
Each file is drafted by an AI analyst, briefed to work like an intelligence officer and told not to take the host’s word for anything. The host reviews it before it is published. The steps are the same for every file.
- The episode is recorded. It is unscripted and is not edited afterwards.
- The full transcript is read. All of it, start to finish. Nothing is taken from show notes or a summary.
- The claims are drawn out. The file sets out what was said, and separates what a source saw from what they worked out or expect.
- Public facts are checked. Dates, figures and events are tested against published sources. Where the episode and the published record differ, the file uses the published record.
- Questions are put to the source. The analyst asks what it would take to counter each problem. The answers become the countermeasures. Guests are asked too.
- Source and evidence are graded. By the rules on this page, which apply to the host as they apply to a guest.
- The file is reviewed and published. The host reads it and can challenge it. Guests are sent their file. Corrections are made where either shows the file is wrong.
The recording is linked from each file as its intercept, so that a reader can check the file against what was said.
Sources
The source is Simon Delaney, whose background is lead generation, identity, verification and software for data processing. Where an episode has a guest, both are named throughout the file.
Checks
Claims about public events, figures and dates are checked against published sources: court records, regulators, industry bodies, standards bodies and established reporting. Where an episode and the published record differ, the file uses the published record.
The sources used are listed at the foot of each file under Exhibits.
The source’s reasoning is tested as well. The file separates what was observed from what was inferred, and marks any step the evidence does not support.
Status labels
Each finding inside a file carries a status.
- Corroborated. Confirmed by a published source.
- First-hand. Observed by the source in his own work. Figures recalled from memory are marked as such.
- Commentary. Published by third parties, some with a commercial interest. Consistent with a finding. Not proof of it.
- Hearsay. Passed on by others and not verified.
- Hypothetical. Reasoning about what comes next. Not a record of what has happened.
- Assessment. The source’s judgement, set apart from what was observed.
Source and evidence
Each file is graded twice, and the two grades are kept apart. One is about the person. The other is about what can be shown. Intelligence services grade reports this way, because a good source and a proven fact are different things.
Source
Was the person placed to see what they describe?
- High. They did the work themselves: ran the test, handled the traffic, reviewed the flow.
- Medium. They work close to it and know how it operates, without having handled it.
- Low. They were not in a position to have seen it.
The grade rests on a stated fact about what the person did. It is a grade of position, not of character. No file vouches for a source, the host included.
Evidence
What can the file point to?
- Published. On the public record: a court record, a regulator, a standard, a report.
- On file. Provided to The Verification Files and inspected.
- Stated. Said by the source and not yet shown. The file says “Contact source for evidence”. Whether to provide it is the source’s decision.
- None. Nothing to point to.
Stated does not mean doubted. A claim from a highly placed source, with the evidence not yet shown, is probable. It is what most first-hand reporting looks like before anyone asks for the records.
Inference is marked as inference. Figures from memory or estimate are marked as such. Hearsay, where a source repeats what they were told, is reported as said and not relied on. Each evidence line ends by stating what has not been measured or tested.
Providing evidence
Evidence is something the file can inspect: records, an anonymised data sample, a document, a screenshot or a published source. Sources, guests and readers can send it to the Verification Desk. When it is on file, the evidence line changes.
Hypothetical
Some of a file is not a claim about what has happened. It is reasoning about what comes next: a source extending what they have seen to where it leads. That is marked hypothetical and kept apart from the grades. It is not a lower grade. It cannot be evidenced yet, because it has not happened. It is judged on the reasoning, and it is often where a file is most useful.
New findings
A finding is not marked down because nobody has published it before. New findings are often the ones nobody has written up. A new finding with evidence on file is graded high.
Reading
A reading is what the file takes a piece of evidence to show. It is kept separate from the evidence itself.
Countermeasures
Each file ends with countermeasures. They are drawn from the episode and from follow-up questions put to the source. They are practical measures and are not legal advice.
Sources, contributors and commercial interests
The Verification Files is produced by Simon Delaney, founder of Provero and Databowl. Provero sells verification, identity and fraud intelligence services. That is a commercial interest in the subjects these files cover, and it is stated here so that readers can weigh it.
Files draw on first-hand professional experience, original observations, published evidence and interviews with industry specialists. Guest contributions are attributed to the people and organisations involved. Their claims are assessed by the same standards as any other evidence. A guest is a contributor, not an author.
Where an organisation provides a service that can help carry out a countermeasure, it may be listed under Operational resources at the end of that file. These can include Provero, guest organisations or other providers. Each entry states the organisation’s connection to this publication. Listing is not an endorsement and is not evidence that a product works.
On a guest’s file, the guest’s organisation is listed first. Provero is listed there only for something different. Not every countermeasure needs a commercial service, and some files list none.
Findings and grades are set by the evidence, not by whether a conclusion benefits a particular business. The countermeasures name no product. Some can only be carried out with a service of this kind.
Challenge and correction
Files that make contested claims carry a challenge status. Anyone with evidence against a file can present it through the contact form, under “Challenge a finding”.