PDPilot
Financial spreading

Financial spreading software that adds up — every total from its own line items

PDPilot turns a borrower's PDF annual report into a standardized, reconciled spread: balance sheet, P&L walk-down and cash flow on one vocabulary, every figure traceable to the printed line, reviewed under dual control and delivered to your credit systems.

PDPilot › Financial spreading software

What financial spreading is — and why it is still done by hand

Financial spreading is the step in corporate credit where an analyst transfers a borrower's financial statements into the lender's own template: the same line names, the same order and the same sign convention for every company, so that ratios, covenants and ratings can be computed and compared. Every annual review, every new facility and every monitoring cycle starts with it.

It is still largely manual because annual reports do not look alike. Layouts, accounting regimes, languages, unit scales and column orders differ from issuer to issuer, and generic PDF tools return text, not a statement. So analysts key the numbers, re-check the totals and paste the result into a spreadsheet or a spreading module — hours per report, before any credit judgement begins.

PDPilot automates exactly this bounded stage. It does not replace the origination, decisioning or monitoring platform; it feeds them reviewed, structured financials.

What PDPilot automates

From the PDF to a decision-ready spread — one governed cycle with a human confirming the inputs and approving the result.

  1. Statement pages are proposed. On upload the balance sheet, income statement and cash-flow pages are proposed with a confidence level and page thumbnails; the analyst confirms or edits them. The extraction runs on exactly the pages confirmed — nothing is hunted elsewhere in the report.
  2. The statements are read as tables. Labels, the year columns as printed in the header, the unit scale declared on the page and side-by-side layouts are resolved deterministically. Statements that span several pages are read as one.
  3. Lines are mapped to one vocabulary. Every issuer, every year and every regime — European GAAP and IFRS statements in English, German, Dutch, French, Italian, Spanish, Portuguese, Danish, Norwegian, Swedish and Finnish — lands on the same standardized balance sheet and P&L walk-down (Revenue → EBITDA → EBIT → pre-tax profit → Net income). The original lines stay visible under each standardized row.
  4. Totals are computed, then reconciled. Total assets, equity, liabilities and every subtotal are the sum of the line items shown; the printed totals only cross-check them. The P&L walk-down must foot to net income. Any gap is shown as a quality signal, never hidden.
  5. Ratios and an indicative rating follow. Leverage, coverage and cash-flow ratios are derived from the spread and the cash-flow page; a metrics-implied rating on the AAA…CCC ladder gives a consistent screening signal.
  6. Review and approval under dual control. Corrections are visibly marked and reversible, submissions lock the figures, and approval requires a second person. Every step lands in an append-only audit log.
  7. Delivery over the API. The approved spread, ratios, rating and approval state are served as structured JSON — the same figures the reviewer approved — for the origination, monitoring and reporting systems.

Sum of components, never the printed total

The one rule behind every PDPilot figure: a total is the sum of the line items you can see underneath it. The report's printed total is read only to validate that sum. When the two differ, the difference is information — a line was missed, misplaced, or the report itself does not foot — and it is shown, not papered over with the printed number.

This is what makes the spread audit-proof. Any figure in a ratio, a rating or a memo can be expanded to the lines it was built from, and each line links to the exact spot on the source page. Extraction and derivation are deterministic rules, not generated text: the same PDF gives the same spread every time, and an audit or model-validation function can reproduce it.

Accuracy is measured, not asserted. Every change to the extraction engine is gated against a curated reference set of close to four thousand European annual-report years: total assets, total equity, total liabilities, revenue and net income are compared with the printed statements at a 0.1 % tolerance, and a change that regresses a single document does not ship.

Inputs PDPilot accepts

PDF annual reports with a text layer

The standard case: listed and unlisted issuers, consolidated or statutory statements, any of the eleven languages above, any European GAAP or IFRS presentation.

Scanned reports

Image-only pages — register scans, rasterized filings, outlined-font exports — get an OCR text layer and go through the same engine. Rows read from scans carry an OCR tag with the page's recognition confidence.

Structured filings (ESEF / iXBRL)

An ESEF inline-XBRL report or a Companies House filing can be imported instead of a PDF. The filing's tagged facts are summed along its own calculation network; tagged totals validate and are never copied.

Excel template import

A standardized bookkeeping export can be mapped deterministically line by line — no page numbers, no layout reading — with a reconciliation screen for every year in the file.

Built for a regulated credit process

Corrections stay visible

A value edited by hand carries a permanent mark with who changed it and what the engine had read; an added line is marked as added. The engine's value is kept and restorable with one click.

Four-eyes approval

The approver must be a different person from the submitter. Approval stores a permanent snapshot of the figures as approved — reopening later never rewrites what was signed off.

Append-only audit trail

Every mapping change, correction, revert, submission and decision is logged with who, when, what and what it replaced. Entries are never edited or deleted.

Integration, not another silo

Every capability — running an extraction, proposing pages, reading the standardized view, the multi-year issuer series, approvals, the credit memo — is available through a documented HTTP API with an OpenAPI contract and API-key authentication for machine callers. The institution's own workflow engine can drive PDPilot end to end, so the standard case runs straight through and analyst attention goes to the exceptions.

An integration kit for Salesforce-based lending platforms such as nCino is available. PDPilot itself runs as a dedicated, single-tenant instance per institution, in a private environment or on-premises; client data never leaves that instance.

Common questions about spreading with PDPilot

Does an analyst still need to check the result?

Yes — by design. The analyst confirms the statement pages, reads the quality signals and approves under dual control. What disappears is the transcription: keying figures, re-adding totals and pasting into a template.

Which accounting regimes and languages are covered?

European GAAP and IFRS statements in English, German, Dutch, French, Italian, Spanish, Portuguese, Danish, Norwegian, Swedish and Finnish, plus UK GAAP accounts filed as scans. Statements are processed in their original language.

Where do the numbers in a ratio come from?

From the standardized spread on screen, and the cash-flow page you supplied for FFO and free operating cash flow. Every ratio input can be expanded to its line items and each line to the source page.

Is the rating a credit rating?

No. It is an indicative, statements-derived signal on the AAA…CCC ladder, blended from four ratios — Debt/EBITDA, FFO/Debt, EBITDA/Interest and FOCF/Debt — for consistent screening. It is not a substitute for a full credit assessment.

See it on your own annual reports

A live demonstration environment is available on request — bring the PDFs your analysts spread today.