WalterFetch · Research & data engineLive

Research that shows its work.

Most research arrives as a conclusion you are asked to trust. WalterSignal’s arrives with the machine that produced it visible — the sources it swept, the providers it ranked, the grade it gave itself before you saw it.

Status
Live in production
Runs on
Self-hosted infrastructure
Output
CSV · PDF · workbook
Verified
July 2026

21

enrichment providers in the stack

3

of them cost money

4

quality tiers, graded before delivery

12

phases in the intro-map flow

Read from the engine, not the brochure · verified July 2026

How it works

One question, five stages, one artifact.

The same engine runs behind every report, every enriched list, and every intro map. Nothing about it is a black box on purpose.

  1. Stage 01

    A question goes in

    Research starts from a decision, not a topic.

    A market to enter, a vendor to displace, a diligence question to close. The brief fixes the scope and the depth before anything runs, because the depth determines how much of the engine is worth spending on the answer.

    • Scoped brief
    • Fixed depth
    • Named decision
  2. Stage 02

    The sources get swept

    Open sources first, and a lot of them.

    Self-hosted meta-search runs the query across many engines at once without a rate-limited API in the middle. Public filings, hiring boards, code hosts, company databases, and the target's own site all get read. Paid sources stay switched off unless the free pass came back thin.

    • Self-hosted meta-search
    • SEC filings
    • ATS + hiring signals
    • Site + tech-stack scraping
  3. Stage 03

    The waterfall runs

    Twenty-one providers, ordered cheapest-first.

    Each record falls through the provider stack until it is complete. Only three providers in the stack cost money, and a record only reaches them if the free ones failed to answer. A per-record spend cap stops a bad run from becoming a bad invoice.

    • 21 providers
    • 3 that cost money
    • Per-record spend cap
  4. Stage 04

    A judge grades it

    The work is scored before anyone sees it.

    An LLM-as-a-judge harness grades the output against the brief, and every record carries a quality tier. Anything that lands in the bottom tier is reworked or dropped rather than shipped. Each field keeps the source it came from, so a claim can be traced back rather than trusted.

    • GOLD
    • SILVER
    • BRONZE
    • FAILED
  5. Stage 05

    An artifact comes out

    Something a person can act on, not a data dump.

    The same backbone renders several finished formats: written research reports, intro maps that trace the warm path into an account, dossier packs on a person or company, and enriched lists with a source behind every field.

    • Reports
    • Intro maps
    • Dossier packs
    • Enriched lists
Put it to work

The engine is the easy part.

Knowing which question is worth pointing it at is the work. Bring us the decision you are stuck on and we will scope what it takes to answer it.