Business evidence methodology
Version 1.0 · Last reviewed 2026-09-01 · Reviewer: Decised methodology review
Decised is deterministic and versioned where models apply. LLMs do not calculate or alter financial outputs.
What we measure
Registered-business counts and sector mix at local authority district (LAD) level; resident population and year-on-year change; business density (businesses per 1,000 residents); observed UK macro and activity series where loaded in a validated bundle; user facts and assumptions in decision cases.
What we do not measure
Customer demand, footfall, commercial rent, trading revenue, personal success probability, or street-level competition beyond descriptive map layers. Crime, news and venue counts do not enter decision scores.
Evidence types
Every displayed value is one of: user fact, user assumption, unverified evidence, observed benchmark, model estimate, or derived calculation, with type, unit, period, source and geography where applicable.
Geography
City evidence profiles use LAD boundaries. Custom searches may use radius context around a point; district counts do not imply street-level conditions.
Missing data
Missing stays missing. Low confidence widens a range or abstains; it does not push a score up or down. We do not invent midpoints, zeros or averages.
Refresh and versioning
Public city snapshots are frozen JSON with explicit data years and snapshot dates. Saved reports are immutable. No live rehydration from latest data. Engine and model versions are recorded on runs and exports.
Release gates
Research-only outputs (for example register-survival probabilities) remain withheld from UI, API, exports and rankings until an explicit engine release gate passes.
Change log
- 2026-09-01: Public methodology page published for SEO release (v1.0).