#6 How FlixBus predicts when you will travel next with Carlos Andrés Arias Robledo (FlixBus)

2021-09-30 · Source: AI LITERACY - A Podcast about Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Intermediate, extended

Summary

Carlos Andrés Arias Robledo, a Data Scientist at FlixBus, details how the company utilizes AI models to predict future travel demand across its extensive network, which spans 36 countries. Despite owning only one bus, FlixBus operates approximately 400,000 daily long-distance connections by functioning as a technology company that partners with bus providers. Arias Robledo explains that demand forecasting involves decomposing historical data into seasonality (e.g., weekly, monthly, yearly patterns), cycles (e.g., non-fixed seasonal shifts), and trends (growth, stability, decline). For new routes or markets, where historical data is scarce, AI models incorporate external data such as population density and existing transportation flows. He highlights the critical role of scenario planning during unprecedented events like the COVID-19 pandemic, where Google Trends data on coronavirus searches was used to anticipate lockdowns and their impact on demand, enabling faster reactions in subsequent countries. The model-building process emphasizes understanding demand drivers and constraints before coding, and using covariates like holidays for improved accuracy.

Key takeaway

For Product Managers overseeing data-driven services, understanding the multi-faceted approach to demand forecasting is crucial. Your team should prioritize building flexible AI models capable of scenario planning, especially for volatile markets or unprecedented events. This enables rapid adaptation and informed decision-making, moving beyond single-value predictions to a comprehensive view of potential futures. Ensure your data scientists are empowered to integrate diverse data sources and continuously validate model assumptions against real-world outcomes.

Key insights

FlixBus uses AI to forecast travel demand by analyzing historical patterns, external data, and scenario planning for rapid adaptation.

Principles

• Decompose demand into seasonality, cycles, and trends.
• Understand business context before model building.
• Validate attribution models with incrementality tests.

Method

Demand forecasting involves decomposing historical data, identifying covariates like holidays, and building models for various scenarios. Incrementality testing uses A/B comparisons to measure marketing channel effectiveness by forecasting "alternative universes."

In practice

• Monitor Google Trends for early indicators of behavioral shifts.
• Use population data for new market demand predictions.
• Conduct A/B tests to validate marketing channel attribution.

Topics

• Demand Forecasting
• AI Models
• FlixBus Business Model
• Time Series Analysis
• Marketing Incrementality

Best for: Product Manager, Data Scientist, AI Product Manager, AI Student

Related on AIssential

• The Hard Truth About Machine Learning for Amazon FBA Sellers — Amazon FBA demand forecasting with ML is challenging due to data hostility, requiring specialized techniques beyond standard approaches.
• 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.
• How to make a cash flow forecasting app work for other systems — Integrated AI agents transform reactive cash flow management into proactive optimization by unifying disparate data sources.
• Autoregressive Models: Predicting the Future Using the Past — Autoregressive models predict future values by leveraging patterns observed in their own historical data.
• Paper Digest: CIKM 2025 Papers & Highlights — CIKM 2025 research highlights advancements in LLMs, graph learning, and multimodal data, addressing challenges in recommendation, forecasting, and privacy.
• When history fails you, borrow from geography — When historical data fails, propagate real-time demand signals from geographically similar, early-affected markets.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI LITERACY - A Podcast about Artificial Intelligence.