- Data Science
- Machine Learning / AI
Hotel Dynamic Pricing Intelligence Platform
Revenue optimization built on causal inference, not just prediction.
A production-shaped pricing system for a five-hotel portfolio. It forecasts demand, measures how guests actually respond to price, and recommends a rate for every unsold night — with an explanation a revenue manager can argue with.
01
Project overview
- What it does
- A rate recommendation engine for a five-hotel portfolio that covers every unsold night.
- Why I built it
- Predicting demand isn't the same as knowing what happens when you change the price. Pricing needs the causal effect, not just a correlation.
- What it solves
- Naive models learn that high prices coincide with high demand and recommend raising rates when demand is high — confusing cause and effect.
- Who it's for
- Hotel revenue managers and portfolio operators.
02
The problem
Historical booking data is confounded: rates were already set in response to expected demand, so a model trained to predict bookings from price learns the wrong lesson. A useful pricing system has to separate how demand moves on its own from how guests respond to the price itself.
03
The solution
The platform forecasts demand for each hotel and night, estimates guests' price response with causal inference methods, and combines the two to recommend a rate for every unsold night. Each recommendation comes with an explanation a revenue manager can review and challenge.
04
Architecture & workflow
- STEP 01
Data
Bookings and rates, five hotels
- STEP 02
Forecast
Demand per hotel and night
- STEP 03
Causal
Measure price response
- STEP 04
Optimize
Rate for each unsold night
- STEP 05
Explain
Reasoning a revenue manager can argue with
05
Technology stack
06
Project demonstration
07
Key features
Demand forecasting
Night-by-night demand for every property in the portfolio.
Causal price response
Estimates how guests react to price, not just how price correlates with demand.
Nightly rate recommendations
A recommended rate for every unsold night.
Arguable explanations
Recommendations a revenue manager can inspect and push back on.