When an investment committee approves a deal that later becomes a write-off, the post-mortem almost never turns up missing data.
The data was in the room. The operating model sat in folder 4.2. The customer concentration table was on page 48 of the confidential information memorandum. The cohort decay curve was on tab six of the workbook.
The failure was not an absence of information. It was an absence of cross-examination.
Between the sponsoring partner’s conviction and the committee’s momentum, the memo did what memos are written to do: it persuaded. Prose covered the arithmetic. Assumptions went unchallenged, because checking them cell by cell meant weeks of cross-referencing that nobody on the deal team had time to run before the meeting.
General Partners are paid to take market risk. They are not paid to sponsor unaudited arithmetic.
The end-to-end mechanics (compiling a deck, reading the score, escalating to the data room) are in the VC Diligence Protocol. This playbook covers the part that protocol stops short of: the committee meeting itself.
1. Three questions every committee must answer
Large language models and junior associates share a failure mode. Asked to summarize a thesis, both optimize for persuasion. The fragile assumption comes back as a well-built sentence.
LLMs optimize for persuasion. askOdin compiles for physics.
An investment outcome carries market uncertainty that nobody can remove. An investment thesis carries logical structure, and that can be audited. Before the partnership signs a term sheet, the committee should be able to answer three questions:
- Does the logic reconstruct? Do the target’s unit economics actually support the top-line claims in the deck, or do they only sit next to them?
- Where does the narrative contradict the model? Not in general. Which slide, and which cell.
- Is there a paper trail? If the company is in distress in month eighteen, can the managing partner show the LP advisory committee that the thesis was cross-examined, on a structured and repeatable basis, before the wire went out?
If the answer to any of the three is “we believe so,” that is a feeling, not an answer.
2. The meeting protocol: appoint a Lead Sceptic
Consensus is the enemy of diligence. When the meeting is built around a sponsoring partner presenting forty slides, the social pressure in the room runs one way.
The fix is structural. Name an adversary.
| Role | Mandate in the meeting |
|---|---|
| Sponsoring partner | Presents the thesis, the upside and the reason to act now. |
| Lead Sceptic | Opens on the contradictions. Owns the case against, for this deal only. |
| Committee | Deliberates on reasoning that has already been cross-examined. |
The Sceptic’s mandate
The Lead Sceptic is a voting partner whose job, on this one deal, is not to endorse it but to find its weakest assumption. The role rotates. It is a duty, not a personality.
Rather than re-reading the deck, the Sceptic works from the Clarity output:
- RUNE Protocol™ (U.S. Provisional Patent No. 63/948,559) extracts the deck’s claims into a logic graph and scores them across the five scored pillars of the Clarity Framework™: Problem Definition, Solution Logic, Market Evidence, Business Model Physics and The Deal Structure, with three audit checks applied as caps. Claims with nothing underneath them surface as brittle assumptions.
- RAVEN Protocol™ (U.S. Provisional Patent No. 63/994,876) cross-examines what the deck claims against what the operating model says, and reports disagreements rather than reconciling them into a cleaner story.
The architectural mechanics of RAVEN’s triangulation engine are protected under U.S. Provisional Patent No. 63/994,876 and are not publicly disclosed.
The first ten minutes
The Sceptic skips the introductory presentation and opens on the contradictions. For example:
Illustrative. “Slide 14 claims an eight-month customer payback. Tab four of the model has acquisition cost at $18,400 against $1,100 of monthly margin contribution. That is a payback of 16.7 months. What does the deck know that the model doesn’t, and why does the plan assume we halve it within two quarters?”
The posture of the room changes. The sponsoring partner is no longer defending a belief. They are answering for whether their own documents agree with each other.
Findings arrive tiered, and the tier tells the committee what to do with each one. The JUDGE Protocol™ (U.S. Provisional Patent No. 64/017,488) classifies every finding as Critical, Major or Minor:
- Minor. Worth a line in the memo. Not a reason to delay.
- Major. The deal team answers it in writing before the vote.
- Critical. A Kill Shot: the protocol stops evaluating and breaks the circuit. The vote waits until the contradiction is resolved or the thesis changes.
3. The paper trail
An LP can forgive a market risk that did not pay off. It is much harder to forgive diligence nobody can show.
Every Clarity Score is benchmarked against 110,000+ Clarity Scores on public deal data, and comes with a Defensible Audit Log™: a sealed, time-stamped record anchored to SHA-256 fingerprints of the exact documents the audit read, and stamped with the version of the Framework that produced it. Change a byte of the deck and the fingerprint stops matching.
Be clear with your LPs about what that record is not.
- It is not a verdict on the company. askOdin audits reasoning, not truth. It does not predict the outcome or replace the partners’ judgment. It documents that the thesis was cross-examined, against which documents, on which date.
- It is not yet independently immutable. “No edits after issue” is our stated policy rather than a control, and the ledger labels it that way.
What happens to the files. No Training Use is a stated policy: extraction runs against an external model API under a paid-tier agreement that prohibits training use, and a deterministic engine does the analysis. Raw documents are held no longer than 30 days under a documented retention ceiling; automated enforcement of that ceiling is in progress, and the security page says so. What persists is the derived judgment record, which can include identifying details, so erasure rights reach it.
If a portfolio company underperforms in year two, the managing partner does not offer the LPAC a recollection of the diligence discussion. They show what was checked, against what, and when.
4. Calibrate on a write-off first
Do not debut this protocol in a live negotiation with a term sheet on the table.
Start with a loss.
- Pick an investment from a past vintage that ended in a write-off or a severe recapitalization.
- Retrieve the deck (PDF or PPTX) and the operating model (XLSX) the committee had on the day it approved the deal. The originals, not the versions that came later.
- Run them through askOdin Clarity.
The audit completes in under three minutes. Then read it against what actually happened.
If it surfaces the contradiction the committee missed, the argument about protocol is over. If it surfaces nothing the partnership did not already discuss, that is worth knowing too: the loss was a market risk rather than a reasoning failure, and that is a different conversation.
The mandate
Persuasion does not scale. Reasoning you can show does.
Audit the reasoning before you write the check.
Adjacent Resources
- askOdin Clarity — the institutional platform.
- The Clarity Framework — the five scored pillars and three audit checks.
- The VC Diligence Protocol — deck to IC memo, end to end.
- The LP’s Blind Spot — what LPs miss underneath the track record.
- The Ledger — what “sealed” means, and what it does not yet mean.
Venture capital is the last unaudited asset class. askOdin provides the infrastructure to close the gap.