SourceSignal
Methodology

We publish the method because the data depends on it.

Every intelligence record is the product of a documented process — from a specific public source, through deterministic processing, to a classified record with its evidence still attached.

The pipeline

01

Source

The monitoring surface is limited to documented public sources: planning registers, procurement portals, contract publications and official notices. Coverage is documented by source, not asserted as a claim of complete national coverage.

02

Detection

Each approved source is checked on a schedule, and changes are detected when new or updated information appears.

03

Evidence

The original source material is captured and stored alongside the record, so every piece of intelligence can be traced back to the publication it came from.

04

Normalisation

Different sources publish in different formats and terminology. Normalisation converts them into a single consistent record structure.

05

Deduplication

The same event is often published multiple times — in different places, under different names. Repeated information is consolidated into a single record.

06

Classification

Every record is classified by industry, signal type and location, so the dataset can be filtered and matched rather than searched.

07

Relevance

Records are matched against the industries, niches and locations you have selected, so you only see the intelligence that matters to you.

08

Delivery

Matched records are made available through the dataset once the platform opens to waitlist members.

Scale, illustrated

0

Sources

0

Signals

0

New records

Illustrative demo metrics only — production figures will be published once the dataset is live.

Evidence over assertion

A record without a source is a claim. SourceSignal keeps the original source publication attached to every record, captured at detection time — so the intelligence can always be verified, and coverage can always be questioned.

Reliability by design

Processing is deterministic and infrastructure-first. That keeps operating costs predictable and results reproducible — the same source in, the same record out.

The next section of the method goes live with the product — the source directory, detection cadence and classification schemas will be public.

Join the waitlist