A source. A reason. A conclusion you can explain.
Frameworks organize the work. Evidence supports the premises. Your reasoning connects them to the conclusion.
From question to reviewable conclusion
- 01
Frame one useful question.
Name the company, the decision, the scope, and the time horizon before collecting material.
- 02
Resolve the entity.
Confirm that filings, websites, and datasets refer to the same company. This matters most for subsidiaries, private companies, and names shared by several businesses.
- 03
Build a source pack.
Prefer direct filings, regulator records, company materials for attributed company claims, and independent sources for external corroboration.
- 04
Separate claim from support.
Record what the source says, where it says it, and whether it supports, contradicts, or only gives context to the claim.
- 05
Show the reasoning.
Give calculations their formula and inputs. Give inferences their premises, mechanism, assumptions, and a condition that would change the conclusion.
- 06
Connect the frameworks.
PESTEL and Five Forces examine the external setting; VRIO examines internal capability; SWOT synthesizes those findings into TOWS choices.
- 07
Challenge the answer.
State the strongest counterargument, remaining unknowns, and the observable evidence that would trigger a review.
What you are saying
Claim kind describes the analytical role of a sentence. It does not prove how well that sentence is supported.
- Fact
- A source-reported observation for a named entity, period, scope, and unit.
- Calculation
- A reproducible result with visible inputs, formula, period, unit, and rounding.
- Estimate
- A range or point estimate that depends on explicit assumptions or a model.
- Interpretation
- A reasoned reading, causal explanation, or strategic implication from supported premises.
- Judgment
- An explicit analyst choice, weighting, threshold, or prioritization.
- Recommendation
- An action tied to findings, decision criteria, tradeoffs, ownership, and a review trigger.
How the claim is supported
- Verified
- Direct eligible evidence supports the claim as worded, scoped, and dated. In the Free Manual interface, a user-attested Verified relationship may be presented as User-checked; that presentation is not independent editorial review.
- Inference
- Verified premises plus visible reasoning support an interpretation, estimate, judgment, or recommendation.
- Unknown
- The evidence is missing, conflicting, stale, inaccessible, or too weak for the claim; show the reason and next research question.
Evidence state remains separate from human-review provenance. Verified does not mean that a human editor approved the page.
Visibility is a separate gate
- Private
- Available only in an authenticated or unguessable workspace; any HTML route uses noindex,nofollow.
- Draft
- Preview or protected editorial work; noindex,nofollow, never in the sitemap.
- Reviewed, noindex
- A public QA preview with a review record and one or more open release gates; noindex,follow, never in the sitemap.
- Indexable
- An individually approved public canonical; only Task 00 can authorize this state after every gate passes.
- Stale, noindex
- A previously public page held for material review; noindex,follow and removed from recommendation modules and the sitemap.
You stay responsible for the work.
The Free Manual workspace structures the analysis you write. AI Research is a separately gated beta unavailable in the public Free Manual route until customer access is authorized. When an example or draft uses AI assistance, the page names the assistance actually evidenced, its evidence cutoff, analysis confidence, limitations, and human-review provenance. Generated text does not become Verified because it is complete or fluent.
Open the underlying sources, follow your institution’s rules for AI and citation, write the argument in your own words, and keep a record of the evidence you relied on.
What a public analysis must show
An indexable analysis requires an unambiguous entity and question, adjacent support for material claims, reproducible calculations, company-specific findings, a counterargument, limitations, a visible AI disclosure, a recorded human editorial decision, an accountable publisher, a working corrections route, matching metadata, accessibility checks, analytics checks, and final release approval.
The Apple example remains an AI-assisted research draft with no recorded human editorial approval. It stays noindex.
Inspect the worked Apple example