Back to all case studies
Hospitality

Driving upsell revenue with federated learning and choice models

Delivered a privacy-preserving upsell engine that personalizes upgrade offers and optimizes pricing without centralizing guest data.

Conversion lift

+18%

Properties federated

65

Upgrade models retrained

Daily

Overview

Hotels sought to increase upsell revenue without sharing raw guest data across properties or relying on one-size-fits-all promotions.

We designed a federated modeling workflow that learns shared patterns across the portfolio while respecting local preferences and privacy constraints.

Challenges

  • Legacy upsell rules ignored traveler context, leading to low acceptance rates.
  • Privacy policies limited the ability to consolidate guest-level data for centralized training.
  • Revenue leaders needed interpretable signals about willingness to pay for specific amenities.

Approach

  • Feature extraction from rich content

    Parsed room descriptions and booking attributes to engineer structured features that explain upgrade value.

  • Elastic-net and multinomial choice modeling

    Combined regularized logistic regression with multinomial logit frameworks to balance interpretability, performance, and real-time pricing control.

  • Hierarchical federated training

    Orchestrated federated averaging with hierarchical Bayesian updates so hotels share statistical strength without exposing raw records.

Impact delivered

  • Delivered personalized upsell offers that adapt to guest personas and booking context in milliseconds.
  • Raised upgrade acceptance while giving management quantified willingness-to-pay metrics by feature.
  • Proved federated learning viable for hospitality AI, balancing privacy compliance with predictive power.

Key lessons

  • Combining federated learning with hierarchical priors unlocks portfolio-wide insights without violating data governance.
  • Display-order optimization is as critical as price selection for upsell success.
  • Elastic regularization keeps models stable and interpretable in fast-changing commercial environments.

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