LeadRadar Book a 4-week pilot
Blog · Challenge coverage

How LeadRadar answers the Orange Systems brief, point by point.

GigaHack 202627 Sep 2026·5 min read

The challenge asks for an AI-powered B2B sales intelligence platform where a sales team defines its own business questions, monitors public sources, detects buying signals, scores companies and prioritises outreach on evidence. Below is the brief's solution scope and expected stack on the left, and the six judging criteria on the right, each with the part of LeadRadar that answers it. Every claim here links to a screen, a formula or a document in the whitepaper; nothing is promised for later without saying so.

Solution scope & stack

✓
Configurable ICP, signal questions, weights, negative rules, scoring logicSignal Builder, per-service templates, simulator, reviewed versions
✓
Data ingestion from company websites, newsrooms, annual reports, news (GDELT / NewsAPI), career pages and job boardsTiered by access status; Crunchbase planned under licence; LinkedIn optional and manual only
✓
Extraction and normalisation from heterogeneous sourcesSchema-validated JSON, exact-quote validation, identity resolution by domain and CUI, deduplication
✓
Dynamic prospect scoring by readiness, with explainabilityRule-based, deterministic P = f(F, R, N) plus a heuristic prediction layer
✓
Signal detection: incidents, leadership changes, tech stack, compliance eventsIncident, new CIO/CISO, cloud migration and NIS2/DORA programme questions ship in the templates
○
Optional outreach generationEmail and call drafts mapped to company context; human-approved, never sent by the platform
✓
React dashboard · PostgreSQL · LLM extraction · HubSpot hooksNext.js / React, Supabase Postgres with row-level security, worker queue, HubSpot preview and commit

Judging criteria, and where to look

25%
Signal relevance & accuracy
Evidence with quote, source and date; Precision@20 ≥ 80% pilot threshold; accept/reject logged from day one
20%
AI/ML innovation
LLMs extract and explain, arithmetic scores; signal sequences → stage and window; counterfactuals and simulation
20%
Configurability
Cold start from a URL; questions, weights, rules and templates edited without code
15%
Usability & UX
A pipeline you read left to right, numbered decisions, plain-language reasons; built for sales teams, not AI experts
10%
Business impact & scalability
Supplier-agnostic twin; new market = new sources and a template; 4-week paid pilot with fixed thresholds
10%
Technical execution
Decoupled ingestion, idempotent CRM writes, tenant isolation, audit trail per decision

What we deliberately left out of the demo: multi-CRM connectors, an MCP assistant and the extended source catalogue are on the roadmap slide, not in code, following the "one thing, done excellently" rule from the demo pack: one service (SCUT NIS2), one CRM (HubSpot), one accounting connector. Time-saved figures are labelled hypotheses until the four-week pilot measures them.

← Back to the landing page