← All deliverables

MTM

Portuguese energy market competitive intelligence — real-time tracking, ML forecasting, copycat detection

Kilo + Peter Shipping May 2026

What It Is

MTM is a competitive intelligence dashboard for the Portuguese energy retail market. It tracks 7 competitors, 318+ offers, and uses machine learning to detect reaction patterns, forecast discount trends, and identify copycat behavior — all in real-time.

The project started as a personal need: working at MEO Energia, I wanted to know what competitors were doing before they knew I was watching. The result is a Streamlit dashboard that combines web scraping, Wayback Machine archives, and ML models into actionable intelligence.

Live Metrics (v0.1)

318
Offers Tracked
7
Competitors
15%
Median Discount
68%
Copycat Similarity

Capabilities

📊 5 Dashboard Views

Overview (KPIs), Competitors (reaction lag), Timeline (discount evolution), Patterns (PCA positioning), Forecasts (time series prediction).

🎯 Copycat Detection

Identifies when competitors copy offers with 68% similarity threshold. GoldEnergy → Galp detected with 3-day lag.

⏱️ Reaction Lag Analysis

Measures how fast competitors respond to market changes. Range: 1 day (fastest) to 12 days (slowest).

🔮 ML Forecasting

Linear regression predicts 8.63 offers/day trend. Per-competitor discount forecasting with 95% confidence intervals.

📡 Real-Time Alerts

Visualping integration monitors 10 competitor pages. Instant notifications when new offers launch.

🗄️ Wayback Archive

66 historical snapshots (Jan–May 2026) for trend analysis. See how pricing evolved over time.

Stack

Python 3.11 Streamlit Plotly scikit-learn SQLAlchemy Alembic PostgreSQL Visualping

Build Log

January 2026
v0.1 — Data Collection
Started scraping competitor websites and Wayback Machine archives. Discovered 66 historical snapshots. Built the first PostgreSQL schema for offers, competitors, and products.
March 2026
v0.2 — Analytics Engine
Added ML models: PCA for competitor positioning, K-Means for offer clustering, Isolation Forest for anomaly detection, Linear Regression for forecasting. The "aha" moment: GoldEnergy copies Galp with 68% similarity and a 3-day lag.
May 2026
v0.3 — Dashboard + Alerts
Built Streamlit dashboard with 5 views and 17 Plotly chart types. Added Visualping for real-time alerts. 318 offers indexed. Dashboard deployed internally at MEO Energia.

What's Next

v0.4
Real-Time Alerts
Email/Slack notifications when competitors launch offers. Trigger-based alerts for price drops and new product launches.
v0.5
Automated Recommendations
"Launch dual-fuel offer now" — ML-powered suggestions based on competitor gaps and market timing.
v1.0
API + Mobile
REST API for external integrations. Mobile-responsive view. Open-source release for other markets.
View on GitHub