Founder Investigation
The record, read closely
Home › Methodology
Editorial

Methodology

How we decide what a story actually shows, and where we draw the line between what is established and what is merely repeated.


Reputation reporting fails in a predictable way: it treats repetition as evidence. A claim gets copied across enough pages that it starts to feel settled, and the fact that no court, regulator or primary document ever established it disappears under the weight of the echo. Our method is built to resist that specific failure. Four principles do most of the work.

Primary records outrank aggregations

We start from documents that establish facts: court judgments and filings, regulator actions and orders, insolvency and administration records, and official company registries. Reporting from named, dated, accountable outlets sits alongside these as context and early warning. Aggregators, reputation scores and automated adverse-media pages sit at the bottom, useful only as pointers toward a primary source. When an aggregation and a primary record disagree, the primary record wins, every time.

Attribution follows the control timeline

The most common error we correct is misattribution across time. A company's name is bound to its founder for the company's whole life, including years after that founder sold control and left. So when something goes wrong, we reconstruct who actually held ownership, board seats and management authority at the moment the relevant decisions were made. If control had already passed to others, the event belongs to them, whatever the search results imply by putting two names side by side.

The ladder of certainty

We keep four words strictly apart, and we never let a story climb the ladder without the evidence to justify it:

Most reputational damage online comes from presenting a rung as if it were a higher one: an allegation dressed as a finding, a corporate settlement dressed as personal guilt. We label each claim at its true rung.

Why name-association scores mislead

Automated risk systems tend to score the proximity between a name and a negative keyword rather than any verified conduct, and they rarely model control timelines. The result is a structural false positive: a founder whose name appears near a later collapse is scored as high-risk even when the record shows they had exited long before. We treat these scores as noise to be checked against the record, never as findings. You can see the full method applied in the Soar Aviation record check.

Sourcing and right of reply

We cite the primary record where it exists and prefer reporting in clusters over any single article. We correct errors that are supported by primary documents, and we log them openly on our corrections page. Subjects and their representatives can respond to anything we publish by writing to the editorial address on our about page.