Breadcrumb — An AI Market Research Platform That Finds Every Buyer in an Industry

The Brief

The client manufactures custom equipment for food-production lines. Their buyers are industrial manufacturers — but nobody sells a clean list of them. The companies hide behind parent corporations, holding groups, and brand names; the facilities that would actually house the equipment aren't listed anywhere; and the people who sign purchase orders don't answer info@ addresses. Sales prospecting meant weeks of manual research per territory, and the picture was stale before it was finished.

Project Type

Custom AI market research and lead intelligence platform.

Domain

Industrial B2B — finding every manufacturer that can use the client's custom equipment.

Data Sources

Industry directories, OCR'd trade publications, OpenStreetMap, enrichment APIs, company websites, news feeds, job postings.

Architecture

Scraper fleet → canonical deduplicated store → AI enrichment → scoring → web portal, all driven by one background job system.

Freshness

Recurring news sweeps, hiring scans, domain health checks, and re-scoring keep the dataset current.

Deliverable

A living web portal — scored leads, company dossiers, maps, and a prioritized call queue — plus CSV exports.

The Problem

Purchased lead lists fail industrial sales teams in three predictable ways. They're organized by legal entity, not by who actually buys — a subsidiary running two plants under a foreign parent shows up as noise, or not at all. They say nothing about facilities — and for equipment sales, the facility is the opportunity. And they're static — a list bought in January says nothing about the company that announced a plant expansion in March.

The client needed something different: a system that could find every company in the market from scattered, messy sources; untangle the corporate structures to reveal who owns what; research each facility down to the building footprint; identify the decision-makers with valid contact details; and keep watching — news, hiring, expansions — so the sales team always knows who to call next, and why.

The Solution

ThinkGenius built Breadcrumb — a market intelligence platform that treats an entire industry as one continuously-refreshed dataset. Scrapers and discovery jobs pull raw company sightings from industry directories, OCR'd trade publications, map data, and enrichment APIs. A deduplication pipeline collapses those sightings into one canonical record per real-world company while preserving full provenance — every source a company was ever seen in stays on the record. AI deep-search enrichment then builds a cited dossier for each company, a scoring model ranks them against the client's ideal customer profile, and a web portal turns the result into a daily sales instrument.

Company Dossiers With Verified Sources

The center of the platform is the company page: an AI-researched dossier that reads like an analyst wrote it. What the company makes, where it manufactures, who owns it, estimated revenue and headcount, and — line by line — which of the client's machines fit which of its production lines. Below every profile sits the sources list: the exact URLs the research pass drew from. When the AI says a company runs 28 plants across three continents under a private-capital parent, every one of those claims links back to where it was found.

Breaking Down Corporate Structures

Industrial markets are full of camouflage: a billion-dollar manufacturer selling through eleven brand names, operating nineteen companies across three continents, held by a private-capital parent. Breadcrumb's entity discovery maps the whole family — parent, sisters, operating companies, brands, and historic names — and cross-references which relatives are already in the system. Above it, the facilities table lists every plant worldwide with its role, per-facility fit score, and estimated square footage from building-footprint data.

The Market on a Map

Territory planning happens visually. The pin map plots every qualified company headquarters and plant across the continent — clustered, tier-filterable, and toggleable between leads and all facilities. A sales trip through a region starts with one look at where the density is.

Adaptable to Any Industry

Nothing in the architecture knows about food production. The platform is three swappable layers on top of industry-agnostic machinery:

• Sources — which directories, publications, map queries, and APIs define the market. Swapping industries means swapping this list.
• Ideal customer profile — the scoring model's definition of a great buyer: product categories, revenue bands, facility types, target roles.
• Fit logic — what the AI enrichment looks for when it reads a company: which of your products map to which of their operations.

The same system that maps industrial manufacturers for an equipment maker could map logistics operators for a warehouse-automation vendor, medical practices for a device company, or contractors for a building-materials brand. Discovery, deduplication with provenance, cited AI enrichment, corporate structure mapping, facility research, contact discovery, and freshness pipelines carry over unchanged.

The Stack

• Python 3.12
• FastAPI + Uvicorn
• MySQL 8 · SQLAlchemy · Alembic
• Playwright (async)
• Tesseract OCR + PyMuPDF
• LLM research & structuring passes
• Contact enrichment APIs (credit-gated)
• OpenStreetMap Overpass + Overture Maps
• News RSS monitoring
• Leaflet interactive & satellite maps
• Server-rendered portal + background job runner
• CSV exports

Why It Matters

Most companies buy market data; very few own their market's data. The difference compounds. A purchased list decays from the day it arrives — companies rebrand, get acquired, open plants, change managers. A platform like Breadcrumb watches all of it: every record traceable to its sources, every score explainable, every signal feeding tomorrow's call queue. The sales team stopped researching and started calling — and the dataset underneath them gets better every week instead of worse.