Methodology

How Finautor keeps answers accurate, and how each product is built.

"Methodology" means something a little different for each product. This page maps the three, and covers in detail how Finautor AI keeps answers accurate.

• Finautor AI — the accuracy safeguards below.
• Signals — how the regime scores are built (the approach, not the formula): Signals methodology.
• The umbrella — how Finautor approaches AI in finance overall: How Finautor approaches AI in finance.

How Finautor AI keeps answers accurate

Trust in financial data is non-negotiable, so Finautor AI is built with layered safeguards that keep answers accurate, honest, and traceable.

Critical numbers come from your data, not the AI

For figures where exactness matters — portfolio equity, P/L, trade summaries, returns — Finautor pulls the value directly from your data and presents it as computed by trusted code. The AI's job is to explain the number, not to paraphrase it.

Accuracy rules baked into every answer

A strict set of rules is built into the AI and can't be overridden by a clever prompt. Each was added in response to a real failure mode:

• Time windows always match — the heading, the tool that fetched the data, and the numbers all refer to the same period.
• Only real fields appear — if a source returns no per-year breakdown, Finautor won't invent one.
• No disguised estimates — "estimated" is never used to launder a made-up number into something that looks real.
• Multiple sources stay labelled — conflicting feeds are shown separately, not blended into one misleading figure.
• If something's missing, Finautor says so — rather than fudging an answer.

A confidence signal on data-driven answers

Each data-driven answer carries a green / amber / red confidence panel you can expand. Green means the underlying data is strong; amber means it's worth a sanity check; red means treat it as a starting point. Expanding the panel explains, in plain language, why an answer scored the way it did.

Sources, citations, and control

Every fact comes with a citation — which feed, filing, or calculation it came from. For generated spreadsheets, Finautor recalculates the formulas so the values you see are really computed. And Finautor never changes your dashboards, Brain, or portfolio on its own — it acts only when you ask, and memory changes can be undone.

How Signals differ

A Signals reading is systematic, not conversational — the same number for every subscriber, computed by a fixed model. Its accuracy discipline is different: the score is served firewall-clean (no inputs or formula), always with an as-of date and a staleness flag. See the Signals methodology for the approach.