Ranking Methodology

1) Lifestyle profile as the primary signal

Every recommendation starts with a Traveler Lifestyle profile built from quiz responses and optional preference refinement. Dimensions such as activity intensity, social energy, novelty tolerance, planning rigidity, and comfort expectations are normalized into a comparable range.

Candidate activities receive a fit score against that profile. We apply penalties when an activity strongly conflicts with dimensions that a traveler has explicitly prioritized. This reduces the chance that a high-rated but personally mismatched activity appears at the top.

2) Multi-source candidate collection

Candidate places and experiences are collected from multiple provider layers to improve coverage and reduce single-source bias. We retain source metadata, then normalize categories into planning buckets such as attractions, dining, and shopping.

If one bucket is underrepresented, fallback sourcing is used to improve balance. This prevents plans from collapsing into one category and helps maintain usable variety for real-world itineraries.

3) Day-flow sequencing

After ranking, activities are sequenced with an energy timeline concept. The sequence aims to avoid unrealistic pacing spikes and supports a smoother travel day: focused blocks, transition windows, and recovery moments where needed.

4) Progressive enrichment

Results can be returned in a progressive mode: a fast initial set is shown first, then enriched as additional source data completes. This improves responsiveness while still allowing deeper coverage and better final ranking quality.

5) Quality and trust principles

We prioritize transparent ranking factors, practical route usability, and traveler-specific relevance. We do not guarantee universal best choices for every traveler; we optimize for the best fit for a specific traveler context.