Short answers for credit analysts, reviewers and integration teams evaluating PDPilot.
The report as a PDF and the positions of three statements in the file: balance sheet, income statement and cash-flow statement — proposed automatically on upload for you to confirm. Plus the basics: issuer name, period end, currency, and whether the statements are consolidated or statutory. Note pages for the borrowings and lease notes are optional.
No — the page's position in the PDF file (1 = the first page of the file). The number printed in the report's footer is usually different, often by a dozen pages or more. The upload proposal pre-fills the positions; if a figure comes back blank or strange, a mixed-up page number is the most common reason.
No. The scale note on the statement page is read automatically. How figures are displayed — millions, thousands or as printed — is a separate choice in the interface.
Yes. A statement page without a text layer gets an OCR text layer and is read by the same engine. Rows from scanned pages carry an OCR tag with the page's recognition confidence.
Yes. Search the public ESEF index by LEI or name, or Companies House by number or name, and import the filing; or drop a saved inline-XBRL, ESEF package or xBRL-JSON file. The tagged facts become the same standardized spread, ratios, rating and memo.
Because every total is added up from the line items you can see — never copied from the printed total. A tiny gap is rounding. A larger gap means a line was missed or misplaced, and that is shown as a finding, not hidden. Expand the rows to see what was captured, then fix it with a re-mapping, a correction or an added line.
Statements are standardized to one fixed vocabulary so that issuers and years are directly comparable. Nothing is lost: expand any row to see the original lines exactly as printed, and click the page link to see the spot on the PDF page.
Not necessarily. The flag means the P&L walk-down (revenue to net income) could not be fully reconciled from the captured lines — usually one line was missed in an unusual layout. Verify against the page and add or correct the missing line; the flag clears when the walk-down adds up.
No. Extraction and derivation are deterministic, verifiable rules. Every figure is computed, traceable and reproducible rather than generated.
The correction is marked permanently with who made it and what the engine originally read. The engine's value is kept and can be restored with one click; the audit log records every step. Corrections belong to that one document — they do not change other years, other issuers or how the engine reads the next report.
They are re-applied wherever the freshly extracted lines still match. Any correction that no longer fits is listed at the top of the document to re-do or discard — nothing is silently dropped.
Any user with the reviewer or admin role — except the person who submitted it. The four-eyes rule is enforced by the platform.
The figures as submitted: a permanent snapshot is stored at approval. If the document is later reopened and changed, the approved snapshot remains viewable exactly as it was approved, and by whom.
From four ratios — Debt/EBITDA, FFO/Debt, EBITDA/Interest cover and FOCF/Debt — blended into a score and mapped to the AAA…CCC ladder, with overlays (net-cash issuers cap at AA; low extraction confidence costs a notch). It is a simplified, methodology-inspired screening signal, not a credit rating and not a replacement for a full credit assessment.
Yes. Saving a correction recomputes the ratios and the rating immediately, so it always reflects the figures on screen.
All analyzed years of an issuer side by side — balance-sheet and P&L series plus the derived ratios — computed live from the same line items the documents show. If the years mix consolidated and statutory figures, or currencies, the view warns prominently.
Yes. Every mapping change, value correction, added line, revert, submission and approval decision is written to a permanent log — who, when, what, and what it replaced. Audit entries are never edited or deleted. Administrative actions are logged the same way.
Two-step sign-in: password plus the current code from an authenticator app, with recovery codes. Roles are analyst, reviewer and admin; higher roles include the lower ones.
In a dedicated, single-tenant instance per institution — in a private environment or on-premises. Client data is processed within that instance and is never blended across clients.
Over a documented HTTP API with an OpenAPI contract: extraction, page proposal, the standardized view, issuer series, approvals and the credit memo. An integration kit for Salesforce-based lending platforms such as nCino is available.
A live demonstration environment is available on request — bring the PDFs your analysts spread today.