METHODOLOGY
How the work gets done — and why it holds up.
Our methodology exists because most research, from the largest sell-side firms to the smallest boutiques, lives partly in the analyst's head. That's an increasingly expensive place for it to live.
Our standing principles
-
If we can't cite where a number, fact, or assertion came from, we don't publish it. AI outputs are sourced like any other input — they don't get a pass for being convenient.
-
The reasoning behind a recommendation is written down before the recommendation is delivered. A client should never have to ask "why does this conclusion follow."
-
Our work product is designed for the PM, not the analyst who built it. The technical complexity stays on our side.
-
Compressed-time research isn't a separate process — it's the same process, run faster, with a different scope. We design for the moment something hits the book, not for the publication cycle.
-
A tool, a report, or an analysis is incomplete if it doesn't come with the documentation that makes it auditable, defensible, and transferable.
What's in every deliverable.
Every piece of work we hand back includes the following:
A clearly stated scope and the assumptions it rests on.
Source citations for every data point and material assertion.
Methodology notes covering every analytical choice.
An "as of" timestamp marking when the analysis was completed.
A version number, for clients who want to track how analysis evolves.
Confidence calibration on every conclusion (high / medium / low).
The model, prompts, and parameters used for any AI-assisted analysis.
A reviewer line — who at Facti Machina signed off.
Sourcing standards
We work with three categories of sources: primary (filings, transcripts, direct communications with company personnel where appropriate), secondary (sell-side research, trade publications, regulatory analyses), and inferred (signals derived from data we don't have direct access to). Each is cited differently in our work, and each is treated with a different level of confidence.
We do not publish anything based on a source we cannot verify. When we use AI tools to assist analysis, the underlying sources and the AI's role in the analysis are both labeled. We do not present AI synthesis as if it were primary research — that practice is increasingly the focus of SEC examination and we treat it as both a compliance and a craft failure.
Methodology
Documentation and 28(e).
On documentation.
The discipline of writing down every methodology choice serves two audiences: your CCO and your future self. Both will eventually want to understand how a specific conclusion was reached. Most research operations fail one or both of those audiences. Ours is built not to.
Your CCO.
Audit defensibility. A complete chain from question to conclusion, with sources and reasoning preserved on the record.
Your future self.
Decision archaeology. The ability to revisit a conclusion six months later and reconstruct exactly why it was reached.
On 28(e) eligibility.
Our research deliverables, sector and security analyses, and methodology documentation are structured to qualify for soft dollar payment under Section 28(e) where the client, in good faith, determines they do. We provide the documentation your broker needs for their records.
We work with each client's broker individually to structure engagements appropriately. We do not opine on whether a specific engagement qualifies for any specific client — that determination belongs to you.
We structure
the deliverable
We document
for your broker's records
You determine
good-faith eligibility
What we don’t do.
▪ We do not provide investment advice and do not manage client assets
▪ We do not publish recurring research notes for general subscription
▪ We do not work with retail-investor-facing firms
▪ We do not work with crypto-only firms or any firm whose regulatory situation would create unmanageable compliance overhead
▪ We do not do AI consulting for non-investment use cases — that work belongs to a different practice
Methodology attribution
This methodology document is reviewed and updated quarterly. Current revision: 22 May 2026. For inquiries about methodology specifics — sourcing rules, the templates that produce our deliverables, the design of our event response system — please contact us directly. We do not distribute the underlying methodology documents publicly.