We find transactions before they come to market.

The non-marketed market is the set of companies that will have to transact within two years and have told nobody. No bank is running a process, so there is no teaser and no data room. The maturity date is filed with the SEC, and the lender who set it has known it since the day the loan was signed.

If you own one of those companies, you are negotiating against the only counterparty that has been preparing for the date, and the extension offer you receive prices the absence of alternatives. If you run a sponsor or a fund waiting for marketed deal flow, you see each opportunity at the same moment as every competitor, in an auction that sets the price. If you are building a platform by acquisition, you are doing that repeatedly, and every auction you enter is paid for out of the same thesis.

These are performing companies with a deadline, which puts the work outside distress and outside restructuring. Most origination starts from a list and argues the evidence afterwards. Ours starts from the evidence, and reading it on your behalf, before a process exists, is what we mean by Evidence First Origination.

Business development companies file a schedule of investments with the SEC each quarter, and each line of that schedule is a loan to a private company. Evidence First Origination runs in five stages, from the raw filing through to a written report on a single situation. Alongside the filings, the engine reads intent and behaviour signals drawn from the record itself: the marks lenders put on their own loans, how books shift quarter to quarter, and amendments as they land.

Never bring a name without the evidence that put it there. A name we cannot source stays out of the ranking, and the same test decides which lenders reach the list.

Before you rely on a data-driven origination claim, you would ask the same six questions we would: where the data comes from, how much of it there is, how often it refreshes, how names are matched, what happens to bad rows and what the output leaves out. These are the answers. Every figure below is a count from the engine's own database as it stands, built from filings up to period end 30 June 2026.

Filings show principal, cost, fair value and rate. They do not show EBITDA, leverage or covenants. They arrive about 45 days after quarter end, so the freshest data is 45 to 135 days old, and an extension may already have been agreed. BDCs are one part of private credit: banks, private placements and most club deals do not appear, so a company missing from the data is simply a company we cannot see. Entity resolution is probabilistic and can be wrong.

These limits are printed in every report we send. Stating what the data cannot say is part of the product, and the anatomy below is the published specification that makes it checkable. Every Lender Report we deliver has these six sections, in this order, whatever the mandate.

Before any engagement, we will build a sample Lender Report for a single situation entirely from real public filings, with a filing date and a source against every number, so you can read the document itself before you decide anything about the firm.

Mandates run in more than one market. The engine's deepest coverage is the US private credit market, where lenders disclose their books to the SEC each quarter and the maturity dates can be read directly. Other jurisdictions disclose less, so coverage there is built from what is filed: statutory accounts, company registries and listed-debt disclosures, supplemented by direct outreach against written criteria where the public record covers less.

Wherever an opportunity comes from, the sourcing and the reasoning are shown on the page, including how much of the picture the local record actually supports.

The filings are parsed, the borrowers resolved and the ranking rebuilt whether or not a client has called, and the lender profiles are drawn from filings published years before any particular engagement. A mandate inherits that work on the first day.