Personalizing hotel demand forecasting with dynamic clustering
Designed an AI demand engine that learns evolving guest personas to power personalized pricing, offers, and merchandising.
Forecast accuracy lift
+4%
Active personas tracked
12
Offer personalization coverage
100% stays
Overview
Hospitality revenue teams lacked a way to differentiate between guest segments when forecasting demand and merchandising offers.
We delivered a demand modeling framework that continuously refreshes guest personas and links them to discrete choice models for room and rate selection.
Challenges
- Traditional forecasts treated all customers identically, blurring willingness-to-pay and booking channel nuances.
- Static segmentation could not adapt to changing traveler mix across seasons and campaigns.
- Pricing teams needed interpretable signals to justify personalized offers and pricing recommendations.
Approach
Dynamic persona discovery
Combined random forests, soft clustering, and Gaussian mixture models to group guests by booking context, stay length, and channel attributes.
Choice modeling inside each cluster
Embedded multinomial logit demand models per persona to capture preferences for room types, rate codes, and price levels, including display-order effects.
Iterative EM refresh cycle
Implemented an EM loop that updates clusters as new stays arrive, ensuring forecasts reflect the latest behavior patterns.
Impact delivered
- Improved prediction accuracy for room and rate-code selection by four percentage points over single-cluster baselines.
- Enabled tailored pricing and merchandising that aligns with guest preferences across every stay.
- Provided revenue managers with persona-level insights that feed downstream optimization and marketing systems.
Key lessons
- Demand forecasting gains accuracy when clustering and choice modeling reinforce each other.
- Model refresh cadences must match how quickly guest mix evolves across channels and seasons.
- Interpretable persona definitions build stakeholder trust in AI-driven merchandising.
Ready to transform your data infrastructure?
Let's discuss how we can help you achieve similar results with a tailored approach for your organization.
Get in touch