Every figure proven — or held. Never guessed.

IVI extracts financial figures from private-market fund documents (capital account statements, capital calls, fund reports, feeders). Anyone can promise extraction. We make a guarantee that can be broken in public: every figure we show you is either proven — backed by an evidence trail you can hand to an auditor — or held for review. A wrong figure is never shown with confidence.

That guarantee, not an accuracy percentage, is the product. Scores move between vendors; a proof artifact does not.

How a figure earns “proven”

• Deterministic first, models last. Models read the document; deterministic code decides what it becomes. An LLM can propose a value, never approve one — the save gate is a mathematical stamp no model can widen.
• Identity math on every transaction. Each statement must satisfy the fund-accounting identity (opening balance + contributions − distributions + income − fees ± carry = closing balance) before it saves. A statement that doesn’t tie is not silently corrected.
• Provenance to the page. Every figure traces to its verbatim text, page, and coordinates in the source PDF. Click a number, see the document.
• Auditor conventions, declared. Where a document is ambiguous, we apply the convention a certified auditor expects and record that we did — never the value that merely matches one record.
• No self-approval. The extractor is never its own validator. Verification runs on independent signals, and a value back-solved from a check can never pass that same check.

We check the GP’s math, not just our reading

Most extraction tools ask one question: did we read this correctly? We ask a second: is the document itself correct? Issuers make errors, and a failed check has two possible authors. When we read a document correctly and it still doesn’t tie, that is not a failure — it is a finding, shown to you with the page as evidence. Telling an LP that a GP’s Q3 statement does not tie by €4,300 — and showing the page — is the most valuable thing we produce.

What “held” means — and why it is rare by design

A hold is a claim that our entire capability chain was exhausted, and that claim has to be true. Only four reasons earn it:

• the document contradicts itself beyond tolerance;
• two trusted sources disagree on the same fact;
• a material figure is genuinely absent and underivable;
• the fund/period/investor identity is genuinely undeterminable.

Everything else — a missing figure derivable from the fund’s own history, a convention question, an ambiguous label — is resolved and flagged, not parked. Held documents sit next to verified ones, clearly marked, never hidden.

Measured, on a real portfolio — with the caveats a buyer needs

Our benchmark is not a hand-picked demo set. It is the complete historical portfolio of a real fund-of-funds: 268 unique documents across 49+ funds, all live quarterly reporting, scored against the LP’s own hand-maintained records — every miss adjudicated to document evidence, not averaged away.

• Median field accuracy 90.4% on the canonical corpus at the last two production deployments (hold-out set: 100%).
• The honesty row: at this sample size, statistics cannot prove a ≥90% claim — the confidence interval is wide (roughly 81–95%). That is exactly why we sell the per-figure guarantee rather than a score: the guarantee holds at any sample size, and the score hardens as new client corpora arrive.
• ≤ €0.05 and under 60 seconds per document on the production stack — measured per quarter, with a regression triggering a route review.

Why publish a caveat against ourselves? Because a vendor quoting “97% accurate” without a sample, a method, or a proof artifact is asking you to trust the number. We are asking you to trust the evidence — and to check it.

What this buys you

Figures you can put your name on in front of a partner or an auditor. Fee and carry errors caught before the wire goes out. A record of where every number came from — retraceable from the screen to the source page, end to end, forever.

Try it on your own documents

The free trial runs full extraction and the identity check on your own statements — no credit card, no restrictions. The fastest way to evaluate us is to watch our hold queue, not our marketing.

The long version

This page is the summary. The full write-up — the guarantee in detail, how each figure earns proven, the hold taxonomy, and the benchmark method with every caveat — is published as a versioned document: Methodology, version 1 (Markdown).