| Filename | Latest commit message | Latest commit date |
|---|---|---|
| README.md | ||
Due Diligence Research With Traceable Claims: A Method for Investors and Analysts
Due diligence is research under time pressure with money at stake. An investment memo might make dozens of factual claims about a company, its market, its technology and its team, and any of them could be wrong. The consequences of an untraceable claim in a memo are rarely visible at the time of the decision; they surface months later, when a figure turns out to have come from a misread chart or an outdated press release. This guide describes a due diligence method in which every material claim carries its evidence, so that the memo can be audited quickly by partners, committees or later reviewers. It borrows from evidence-traceable research practice, the same principle behind workbenches such as Yila, and applies it to commercial analysis.
Separate claims from conclusions
An investment memo mixes three kinds of statement:
- Factual claims: revenue last year, number of customers, size of the market, results of a clinical study, patents held.
- Assessments: the team is strong, the moat is defensible, the market is growing.
- Recommendations: invest, pass, invest with conditions.
Assessments and recommendations are judgements, and they should rest on factual claims. The first discipline of traceable due diligence is to make that dependency explicit. For every assessment, list the factual claims that support it. For every factual claim, record the evidence.
Classify evidence by source and verification
Evidence in due diligence comes from very different places. A simple classification:
- Company-provided and verified: data room documents checked against independent sources, audited accounts, contracts reviewed by counsel.
- Company-provided and unverified: management presentations, internal dashboards, projections.
- Independent primary: public filings, regulatory records, patent databases, court records, scientific publications, customer interviews conducted by your team.
- Independent secondary: analyst reports, trade press, market research publications.
- Informal: conversations, social media, unattributed reports.
Every factual claim in the memo should carry its class. A revenue figure from audited accounts and a revenue figure from a pitch deck are not the same claim, even if the number is identical.
Capture evidence as you go
The worst moment to reconstruct sources is the night before an investment committee meeting. Build the evidence record during the work:
- When you extract a number from a document, record the document, the page and the exact wording.
- When you take a claim from a call or interview, record who said it, when and in what context.
- When you rely on a web page, archive it with a date, because company websites and pricing pages change.
- When two sources disagree, record both and note which you used and why.
AI research tools can accelerate the extraction step considerably, especially with long documents such as technical papers, regulatory filings and market reports. The safeguard is traceability: the tool must link each extracted claim to the passage it came from so the analyst can confirm it. Yila's sentence-level evidence links follow this approach, which makes it practical to check a summary of a forty-page technical report in minutes rather than re-reading it.
Test the claims that matter most
Not every claim deserves the same scrutiny. Identify the handful of claims that, if wrong, would change the recommendation. These typically include:
- Revenue, growth and retention figures.
- The size and growth of the addressable market.
- Technical claims that underpin the product's advantage.
- Regulatory status and pending approvals.
- Key customer and partner relationships.
For each, seek independent confirmation. A technical claim might be checked against peer-reviewed publications or an expert interview. A market size might be triangulated across two or three independent estimates. A customer relationship might be confirmed with the customer directly.
Record the outcome of each test: confirmed, partially confirmed, contradicted or unverifiable. A memo that admits "we could not independently verify the claimed pilot results" is more useful than one that silently relies on them.
Research the science and technology carefully
Deep-tech and life-science deals often hinge on scientific claims. Founders cite papers, sometimes accurately and sometimes optimistically. For any scientific claim central to the thesis:
- Read the cited paper, not just its abstract.
- Check whether the result has been replicated or contradicted.
- Note the stage of evidence: in vitro, animal, early human, controlled trial.
- Look for retractions, corrections and critical commentary.
- Check whether the company's product is actually what was tested.
This is literature review work, and the same tools and habits apply. An assistant that reads full texts and keeps claims linked to passages is far safer here than one that summarises abstracts.
Write the memo with visible evidence
The memo itself should make the evidence trail easy to follow without burying the argument. Practical techniques:
- Footnote every material factual claim with its source and evidence class.
- Use consistent confidence language: "verified", "company-reported", "estimated", "unverified".
- Include an appendix listing the key claims, their tests and outcomes.
- Flag explicitly any assessment that rests mainly on unverified evidence.
Committees often skim memos, but a visible evidence structure changes the conversation. Questions shift from "where did this come from?" to "how much weight should we put on this?", which is the conversation that improves decisions.
Post-decision value
A traceable memo continues to pay off after the decision. When the company reports results, you can compare them with the claims in the memo and see which assumptions held. When a deal goes badly, you can identify which claims failed and whether better research would have caught them. Over time, this feedback improves the firm's diligence process.
Summary
Due diligence becomes more reliable when every material claim carries its evidence and its verification status. Separate claims from conclusions, classify evidence, capture sources at the moment of reading, test the claims that matter and write memos that make the trail visible. Tools that keep summaries linked to their source passages, such as the evidence-traceable workbench at https://yila.ai, reduce the cost of doing this well, but the discipline itself is what protects the decision.