Core Engine // 02
Revenue Intelligence
ICP scoring, buying role detection, and pipeline diagnosis. Updated daily.
A machine learning engine that scores every active deal across 12 signals and pushes results back to HubSpot daily.
Deal scoring playground
Deal Scoring Playground
Enter deal details and run the engine to see a real-time ML score with full factor attribution.
How it works
ICP mapping
The engine starts with who actually buys. Industries are tiered by historical win rate — tier 1, tier 2, hidden gems. A lead from the right industry starts with a different score than one from the wrong one.
Signal collection
Each deal is evaluated across 12 signals: industry fit, traffic source, ticket size, buying role of the contact, engagement depth, stage velocity, and pipeline history.
Buying role detection
Talking to an end user without a decision maker is a leading indicator of loss. The engine detects who is in the room and adjusts the score accordingly.
ML prediction
A Random Forest and Gradient Boosting ensemble trained on closed deals produces a 0-100 score with full attribution. Not a number. A diagnosis.
RevOps sync
Scores write back to the CRM daily. Sales teams see updated signals in their existing workflow. No new dashboards, no new logins. The intelligence meets them where they work.
Capabilities
ICP tier scoring
Industries are classified by win rate into tiers. The engine knows which verticals close and weights every deal from that industry accordingly.
Buying role detection
Decision maker, influencer, or end user — the engine reads who is involved and penalizes patterns that historically lead to loss.
Source attribution
Not all leads are equal. The model learns which acquisition channels produce deals that close, not just deals that start.
Ticket size signals
Deal size is a proxy for seriousness and organizational fit. Enterprise deals from the right industry score higher regardless of stage.
Explainable output
Every score comes with a breakdown: which signals are lifting it, which are dragging it down, and what the team should act on.
LangGraph agent
A 7-node reasoning agent reads active deals, generates natural language summaries of risk and opportunity, and flags deals needing immediate attention.
What it solves
RevOps is not a dashboard problem. It is a signal problem. Sales teams are sitting on the data that predicts who will buy — they just cannot see it. This engine surfaces those signals daily, in the tools the team already uses, before the opportunity closes.
Case Studies
ecommerce
Detección Temprana de Oportunidades de Upsell en Ecommerce: Clustering de Comportamiento Producto
El presente estudio analiza la implementación de un modelo de clustering de comp...
8% — Tasa de conversión de productos de upsell
professional-services
Modelos Predictivos de Churn B2B: Ingeniería de Features desde Senales de Uso y CRM para Professional Services
El presente estudio analiza el problema del churn en empresas de servicios profe...
8% — Reducción en Tasa de Churn
education
Forecasting Avanzado del Pipeline: Redes Neuronales Recurrentes vs. Stage-Weighted en EdTech
Este estudio de caso analiza la aplicación de Redes Neuronales Recurrentes (RNN)...
8.5 leads — RMSE en Predicción a 7 días
Work with this engine
Get in touch →