When history fails you, borrow from geography

2026-06-02 · Source: The Airbnb Tech Blog - Medium · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Advanced, long

Summary

Airbnb developed a novel forecasting system to address the challenges of uneven demand recovery during the late 2020 through 2022 COVID-19 period, when traditional models failed due to unprecedented shocks. Instead of waiting for local data, the system "borrows from geography" by propagating observable evidence from early-recovering markets to later-affected ones. This approach leverages the sequential, asynchronous nature of global events like vaccine rollouts and border reopenings, where regions like Europe and North America experienced similar demand compression and recovery phases but on different timelines. The system utilizes a hierarchical Bayesian framework, where the posterior from an early-affected corridor becomes an informative prior for a late-affected one, weighted by structural similarity. This enabled Airbnb to generate timely, informative corridor-level forecasts even with scarce local data, a method now integrated into their standing infrastructure for any sequentially unfolding market changes.

Key takeaway

For Data Scientists or Machine Learning Engineers building forecasting systems for global operations, if your markets experience asynchronous, sequential changes, you should implement a geographic prior propagation strategy. This allows your models to generate timely, informed forecasts for data-scarce regions by leveraging real-time signals from structurally similar, early-affected markets. Consider adopting a hierarchical Bayesian framework to automatically balance propagated priors with accumulating local data, ensuring robust predictions during dynamic periods.

Key insights

When historical data fails, propagate real-time demand signals from geographically similar, early-affected markets.

Principles

• Geographic structure offers underutilized forecasting signals.
• Sequential rollouts provide valuable leading indicators.
• Hierarchical Bayesian models effectively share information.

Method

Use a hierarchical Bayesian framework where an early-affected corridor's posterior updates as an informative prior for a late-affected, structurally similar corridor, balancing propagated signal with local data.

In practice

• Apply to new product feature rollouts across regions.
• Inform forecasts during staggered regulatory changes.
• Predict demand shifts from commodity price shocks.

Topics

• Geographic Forecasting
• Bayesian Hierarchical Models
• Demand Forecasting
• Sequential Rollouts
• Travel Industry Analytics
• COVID-19 Impact

Best for: AI Scientist, Research Scientist, Data Scientist, Machine Learning Engineer, Director of AI/ML

Related on AIssential

• Forecasting as Rendering: A 2D Gaussian Splatting Framework for Time Series Forecasting — Time series forecasting can be fundamentally improved by treating it as 2D generative rendering using anisotropic Gaussian splatting.
• GPS-Enhanced Tourist Mobility Modeling with Seasonal Spatial Priors and LLM-Based Activity Chain Generation — The framework combines GPS data, demographics, and LLMs to simulate realistic, demographically-aligned tourist mobility for urban planning.
• 7 Steps to Mastering Time Series Analysis with Python — Mastering time series analysis requires a sequential, structured approach, from understanding unique data properties to robust deployment.
• Product Forecasting through Time Series Analysis (Modelling) — Effective product sales forecasting requires a hierarchical, multi-model approach combining ML, deep learning, and statistical methods.
• How Earth-2 Delivers Global AI Weather Forecasts on Demand — AI models can simulate atmospheric physics directly from observational data, outperforming traditional methods.
• Geographically-aware Transformer-based Traffic Forecasting for Urban Motorway Digital Twins — Geographical awareness via mutual information significantly enhances Transformer-based traffic forecasting accuracy without added complexity.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Airbnb Tech Blog - Medium.