Independent Sponsor Intelligence

Nothing is stored as a current value that overwrites history. Each thing a source said is an observation with a subject, a predicate, a value, the date it was effective, the date we read it, the source, the URL, a quote where one exists, an evidence class (fact, inference or estimate) and a content hash. Re-reading a source is free and safe: a second sighting of the same fact increments a counter rather than creating a row. Current-state views sit on top and are rebuilt from the tape.

An entity is created when nothing resolves. Resolution is by external identifier first (SEC CIK, file number, CRD, EIN), then by exact normalised name within the same entity type and, where both sides know one, the same state. Nothing is merged on a similar name. Sponsors are inferred from acquisition vehicles: two vehicles that share a named executive officer or promoter are joined at full weight, two that share a filing street address at reduced weight, and known registered-agent addresses are excluded. Every join carries the evidence that produced it and can be undone without touching the tape.

Private company EBITDA is rarely available. The product prints a probable band, never a point: workforce proxy (benefit plan participants, or an observed employee count) multiplied by a sector revenue-per-employee range and a sector EBITDA margin range, with the method and a confidence that falls as the sector's economics widen. Where a QoE-backed figure is observed it replaces the band and says so.

Dealability (size band, tenure, benefit footprint, workforce stability, sector demand, ownership type), sponsor fit (stated demand from an anonymised market prior plus observed vehicle activity in the vertical, size in the demanded band, region), transition readiness (dated signals, each with a weight and a half-life: succession, ownership events, transaction preparation, tenure, market pull), urgency (the freshest strong signal alone) and data confidence (source breadth, freshness, estimate quality). Opportunity is dealability times a transition factor times a confidence factor, and is used for ranking only. No age is ever a component: the system holds no verified age for anyone and will not guess one. A company whose transition score rests on tenure and market pull alone is not "for sale" and the page says so.

A sponsor's buy box is what its vehicles show: the industry groups they file under, the states they are organised in, the equity raised per vehicle (a distribution, not a number), the share flagged as business combinations, the filing dates. A capital provider's mandate is what the regulator publishes (SBIC strategy, fund size, average investment, whether it is making new investments), what its Form ADV reports (private fund assets, fund count, registration date) and what its fund vehicles sold. Sponsor friendliness begins as a class prior labelled inference and is replaced by observed sponsor counterparties as transactions are ingested.

Every sponsor entity carries one status, derived from adjudication and the classifier and checked the same way in the database. Only a verified sponsor firm is listed by default, badged, or shown beside a company, and the database refuses to store a company pair for anything else.

• Verified sponsor. Evidence identifies a firm: a publisher named it as the buyer in an announced transaction, a law firm or capital partner page names it as an independent sponsor, or its own site says so.
• Unverified filing group, research only. A group of SEC Form D vehicles joined by shared principals or a shared address. No evidence yet identifies the firm behind it, so it is never shown beside a company as a sponsor.
• Unverified candidate, research only. One vehicle, one mention or an individual. It is kept for research and is never shown beside a company as a sponsor.
• Unverified, research only. The evidence does not settle whether this is a sponsor firm. It is kept for research and is never shown beside a company as a sponsor.
• Not a sponsor firm. The evidence reads this entity as something other than an independent sponsor firm. It is never shown beside a company as a sponsor.

A capital relationship is a retrieved fact: a provider named in an announcement as financing a sponsor's deal. When a provider has financed two or more observed deals for the same sponsor it is shown as a repeat relationship, in a sentence with the roles, the dates and a link to each announcement. Counts are what DFX has read, mostly in lender and sponsor news archives, so they undercount and are not a forecast.

DFX does not recommend a capital provider. In a strict as-of backtest over 198 first-time financings, the DFX capital model put the actual provider in its top five 22.7% of the time, against 42.6% for a plain count of each provider's prior deals, so the model is not shown anywhere. Repeat relationships are also read with care: of 46 sponsor and provider pairs seen on two or more deals, 42 were seen on a single website, almost always the lender's own deal archive. When a sponsor's next financing was read from a different kind of source, it used its previous lender in 3 of 28 cases; when both came from the same kind of source, in 47 of 72. A repeat relationship is a fact about those deals, not a rate at which sponsors return to a lender.

Every ownership or transition row on the record is placed in exactly one class, and the classes are counted apart. CONFIRMED TRANSACTION: public transaction evidence identifies the ownership transaction and usually the buyer; the URL is on the row. LIKELY OWNERSHIP CHANGE: the plan sponsor of record became a holding or acquisition vehicle and no announcement has been found; the rule that inferred it and its measured precision are on the row (a deal-structure word in the new name was an ownership change in 94% of adjudicated cases, a bare holding word in 70%, over 77 cases on the 2023 and 2024 tapes). COMPANY TRANSITION: something happened to the company or its plan, the cause is not asserted: a final filing then absence from the record, a final filing whose next year is not yet published, a plan ended while the sponsor filed on, a rename, an EIN change. UNRESOLVED RECORD DEPARTURE: the company left the observable Form 5500 universe with no final filing and no successor EIN. Departures are never counted as transactions, never appear in a headline number and never enter a sponsor's history.

A departure is measured strictly: the EIN, and any successor EIN the filing names, is absent from every later plan year's Form 5500 and 5500-SF at any size and in any industry. A sponsor that shrank below the size floor, grew past the ceiling or changed business code is still filing and is not a departure. On the 2022 and 2023 tapes, measured against that raw record, 79 to 80% of sponsors that marked a return final were absent the following plan year, against 1.8 to 2.1% of sponsors that did not (the 88 to 90% quoted earlier was measured against the sized universe only). That is next-year absence after a final filing, not a sale rate, and it is never described as one.

A company survives its own rename. A HoldCo rename or a new EIN is recorded as an event on the same longitudinal company, whose page shows its current name, what it was called before, its prior EIN, the ownership event and the buyer where known, with its workforce and plan history intact.

A research claim is refused without a public URL and a quote. An estimate is never displayed as a fact. A source with no rights determination is ingested but never published. People are never pages; a principal's name appears only where the public filing already prints it.

Early coverage is weighted by an anonymised demand model built from a registration sample of the US independent sponsor ecosystem: aggregate distributions over sectors, sizes, structures and geographies, with every cell below a floor folded into other. It influences what the system looks for and is never cited as evidence about any entity.