How it works

Rules decide. AI explains.

The order matters, and it is enforced in the code rather than promised in marketing copy: nothing a model produces can change a score, an eligibility decision, or a number.

The pipeline

Every result in the product travels this path, in this direction only.

Data

Companies House filings, officers, PSCs, charges — the public record, unaltered.

Structured facts

Figures extracted from filed iXBRL accounts, each carrying its source, period, and confidence.

Deterministic rules

Your Buy Box criteria, evaluated as hard filters, soft preferences, and exclusions.

Deterministic score

A 0–100 fit score with a breakdown that sums to exactly 100.

AI explanation

Plain-language narrative describing a result that was already decided without it.

The scoring and signal engines structurally cannot reach the AI layer or a database client — that boundary is enforced by lint rules in the repository, so a future change that tried to blur it would fail the build rather than ship quietly.

Describe your thesis

Write what you're looking for in plain English — “profitable UK manufacturing businesses doing £1–5m with an owner near retirement” — or build the criteria by hand. If you write it in prose, it's parsed into structured criteria and shown to you for confirmation. Nothing takes effect until you approve it.

Criteria become rules

Each criterion becomes one of three things: a hard filter that a company must pass, a soft preference that contributes points, or an exclusion that rules a company out. Hard filters and exclusions are pass/fail — they are never traded off against a good score elsewhere.

Companies are scored

Matching runs against synced company data. A company that fails a hard filter or hits an exclusion is not “low scoring” — it is ineligible, and shown as such. Everything that passes gets a fit score, a confidence level reflecting how much data was actually available, and a data-completeness breakdown.

You see exactly why

Open any score and you get the arithmetic: which criteria matched, which didn't, what each contributed, and which figures were unknown. Unknown data is never silently treated as zero — it lowers confidence instead of quietly inventing a result.

Changes are tracked

The register keeps moving. New filings, charges, officer changes, and financial deterioration are detected, deduplicated, and ranked by severity, so Opportunity Radar tells you what actually changed rather than replaying everything.

Research, then decide

Shortlist into watchlists, compare side by side, work a diligence checklist prioritised from the company's own weak points, then draft an investment memo and move the deal into your pipeline.

See it against a company you already know

The fastest sanity check is to score a business you understand and read the breakdown.

Get started