Research Standards
Make the work checkable.
If a reader cannot tell where a fact ends and my judgment begins, I have not done the job. These are the rules I use to keep sources, uncertainty, tradeoffs, and later revisions visible — and, below, exactly how a claim gets checked before it's published, and where to go verify any of it yourself.
- 01
Start with primary material.
Company filings, earnings materials, and direct disclosures are the preferred starting point. Secondary reporting adds context but should not quietly become the factual foundation of a thesis.
In practice: every earnings report on this site opens with the company's own 8-K/6-K/10-Q filing and investor-relations release before any secondary source is cited — see the sourcing hierarchy for the exact tiering.
- 02
Label inference as inference.
A reported fact, a model assumption, and a personal judgment do different jobs. The writing should make the boundary clear enough for a reader to disagree intelligently.
In practice: position pages separate a filed number from a personal read rather than blending the two into one sentence — the "Risk" field on each position page is explicitly framed as judgment, not as a reported fact.
- 03
Write the countercase before the outcome.
Each position carries a bear case and, where practical, pre-mortem review triggers. A risk is not a boilerplate disclaimer when it can change the decision.
In practice: every position's bull/base/bear case is written before the next earnings print, not rewritten after the fact to match how the quarter actually went.
- 04
Disclose the position and the limits.
AEA Capital Research is a personal, educational project—not a fund or investment adviser. Holdings, policy constraints, and methodology limits are stated in the disclosures and Investment Policy Statement.
In practice: the Compliance Ledger checks every IPS cap against the live portfolio and lists breaches until they're actually resolved — not until it's convenient to stop mentioning them.
- 05
Revise without erasing the original reasoning.
When evidence changes a view or exposes a process flaw, the better record is an update or post-mortem, not a silent rewrite. The Research Practice page keeps those revisions visible.
In practice: a corrected weight or return figure gets a dated note explaining what changed and why, rather than a quiet overwrite with no trace of the earlier number.
Process
How a claim actually gets checked.
The five rules above are the standard. This is the mechanical version — the four steps a specific number or claim goes through before it appears on the site.
- A
Pull the primary number.
Start from the filing, the brokerage-linked Blossom feed, or the live market-data pipeline — never from memory or a secondary summary, per Standard 01.
- B
Cross-check it against a second source.
Watchlist fundamentals and consensus figures get checked against at least two independent aggregators before publishing; portfolio figures are checked against the Holdings page they're supposed to match.
- C
Write the uncertainty in, not around.
If a figure is stale, unverifiable, or an estimate, the page says so directly — the Yield Curve monitor's "not yet re-verified this cycle" note is the pattern every other page follows.
- D
Date it, and let it be checked again later.
Every live page carries a "Data verified" date. When the number moves, the date moves with it — and the old number doesn't just disappear without a record somewhere it changed.
Verification
Where to check this yourself.
Standards are only real if a reader can go test them. These are the four pages built specifically to be checked against, not taken on faith.
Sources & Research Log
Every citation on the site, aggregated and counted — which source tier backs which kind of number, and how often each one actually gets used.
Standard 04Compliance Ledger
Every Investment Policy Statement cap, checked live against the actual portfolio. Breaches stay listed until resolved, not until convenient.
Standard 05Post-Mortems
Dated write-ups of the calls that missed — the original thesis, what actually happened, and the specific process flaw it exposed.
Standard 04Disclosures
The formal statement of what this site is and isn't, methodology limits, and forward-looking-statement language, in one place.
A note on certainty
Research can improve a decision without making it certain. The standard here is not perfect forecasting; it is disciplined reasoning that recognizes tradeoffs and can be inspected, challenged, and revisited later.