Jobs and employers
Open postings with their full text, the employers behind them, and a changes feed that keeps a copy in sync, with dates a backtest can trust.
BDC booksSEC EDGARBy ticker or CIK
What a business development company reports lending, read from its 10-Q and 10-K filings: who it lends to, on what terms, and what it says each loan is worth. Then how every other lender we read marks the same loan. One lender at a time, every quarter since its filings became structured data.
Hold shares in a BDC? Its schedule of investments is public, and so are its co-lenders'. When three other lenders mark a loan at 85 and yours carries it at 97, that is worth knowing before the next NAV.
$ pdatum bdc request ACME ACME: queued. Follow it with: pdatum bdc book ACME # a lender not read before takes a while $ pdatum bdc pull ACME --out acme filings: 16 rows positions: 3,595 rows companies: 412 rows comparisons: 1,212 rows # and five more, each checked against the manifest acme $ jq -c 'select(.gap >= 10) | {company, ours, median, gap}' \ acme/comparisons.jsonl {"company":"acme-fastening","ours":97.25,"median":85.5,"gap":11.75}
Illustrative: Acme, its ticker, its borrowers and its numbers are invented.
The questions a book is built to answer, and the table that answers each.
comparisons sets this lender's mark beside every other
lender's on the same term loan, quarter by quarter, with the median and
the gap. coverage says how much of the book has a
comparison at all.
events lists every entry, exit and moved field between
consecutive filings, before and after. amendments lists the
term changes lenders report on a shared loan: PIK on or off, a spread
moved, a maturity extended.
A 10-Q for 30 June becomes public weeks later. filings
keeps the date a figure describes apart from the day it was filed, and
recorded_at serves a book exactly as we held it on a past
day.
warn holds the state WARN Act notices that name a
borrower in the book: the layoffs and closures an employer has to
announce ahead of time.
figures holds the numbers a report would quote, already
computed, each with its unit, its date and a definition in words. A
book's caveats are sentences to read before any of
them.
A book is needs_work when the positions read do not sum
to within 0.5% of the total the lender reports. reconciliation
shows both numbers either way.
A manifest and nine tables, as JSON lines. The same
version is the same data, byte for byte, so cache on it; a
book is rebuilt when its lender files. The API
reference has every field.
Money is whole US dollars, except principal, which is in its own currency. A mark is fair value per 100 of par.
How the data behaves, in brief. The front page lists every source we read, and the API reference has the rest.
coverage says how much of a book other lenders hold
too.Open postings with their full text, the employers behind them, and a changes feed that keeps a copy in sync, with dates a backtest can trust.
The contracts and grants federal agencies have awarded any company you name, from USAspending.gov: how many, how much, from which agencies, and when the work ends.
Every private company the BDCs we read hold, under every name their lenders file it as, with one slug. Any name can be looked up; one we do not hold is searched for on EDGAR.
The layoff and closure notices states publish under the WARN Act, read every day and served in one shape: the employer, the place, how many people and when, linked to the companies lenders hold.
Franchisors' registrations and exemption notices with the states, every Texas filing back to 1985, with their brands, dates and status, linked to the companies lenders hold.
Every private capital raise a company has reported to the SEC on Form D since 2008: how much it offered, how much it sold and to how many investors, linked to the companies lenders hold.
Keys are issued by hand while the API is young. Write to pdatum@pearachute.com and say what you are building and which lenders you want read.