Document-First Investigation¶

Why It Matters¶

Human memory is unreliable, people lie or shade truth, and testimony can be retracted or denied. Documents are fixed. An email written in 2015 can't change its story in 2024. Financial records, meeting minutes, internal memos, and contracts document what actually happened at the time, not what people remember or claim. This asymmetry means documents are stronger evidence than testimony, and smart investigation builds outward from documents.

The Key Move¶

Start with the paper trail. What documents must exist if this claim is true? Financial transactions leave bank records and invoices. Corporate decisions leave board minutes and emails. Government actions leave permits, correspondence, and budgets. Request documents first through FOIA, subpoenas, or public records. Analyze documents to identify participants, timelines, and inconsistencies. Then interview sources armed with documentary evidence: "This email says X - can you explain the context?" This reverses the power dynamic. The source can't deny what's documented, and you can detect when they're lying or misremembering by checking against fixed records.

Classic Application¶

Panama Papers investigation. The leak provided 11.5 million documents from Mossack Fonseca law firm. Rather than starting with sources, journalists used documents to map corporate structures, identify beneficial owners, trace money flows, and timeline transactions. Only after document analysis did they approach individuals: "We have documents showing you transferred $X million through Y shell company on Z date. Can you explain this?" Many subjects couldn't deny documented facts and either confirmed details or declined comment, but couldn't offer false alternative narratives.

Surprising Application¶

Product usage analysis. Rather than asking users "what features do you use most?" (unreliable recall), analyze usage logs (documents of actual behavior). Users dramatically mis-report their behavior - they think they use certain features because they value them, when logs show different patterns. Document-first (logs, analytics, error reports) reveals actual usage patterns, then user interviews explain why: "Logs show you never used feature X despite saying it's essential. What's going on?" Often: "Oh, I thought I did, but actually I use Y instead."

Failure Modes¶

Assuming documents are complete (selective retention and destruction happen). Not verifying document authenticity (forgeries exist - check provenance, metadata, and internal consistency). Treating documents as self-explanatory (they often need context from sources who understand technical jargon or organizational culture). Becoming paralyzed waiting for the "smoking gun" document (sometimes testimony is all you have). Not requesting documents because you assume they don't exist or won't be released (bureaucracies often release more than expected if you ask correctly).

Go Deeper¶

Gup, "The Art of Business: Journalism After Watergate" on document-driven investigation. Weinberg, "The Reporter's Handbook" Part II on public records and document analysis. Houston, "Computer-Assisted Reporting" on analyzing large document sets.