Learn by Directing AI
All materials

data-dictionary.md

Data Dictionary: Cumbre Adventures A/B Test

Dataset Overview

Source: Booking platform export from Cumbre Adventures (cumbreadventures.bo) Date range: 60 days (October 1 - November 29, 2025) Total rows: ~4,200 (one row per visitor) Experiment description: A 50/50 random traffic split between the original booking page (version A) and the redesigned booking page (version B). Every visitor to the English-language booking page was randomly assigned to one version.

Column Descriptions

Column Type Description
visitor_id String Unique visitor identifier. Format: "V0001" through "V4200". One row per visitor.
page_version String Which booking page the visitor saw. "A" = original page, "B" = redesigned page. Approximately 50/50 split.
visit_date Date (YYYY-MM-DD) The date the visitor accessed the booking page. Spans the full 60-day experiment period.
tour_selected String or empty Which tour the visitor booked. One of: death_road, huayna_potosi, cordillera_real, paragliding. Empty if the visitor did not book.
booking_completed Boolean Whether the visitor completed a booking. "true" or "false".
booking_value Numeric (USD) The booking value in US dollars. 0 if no booking was completed. Positive value if a booking was made. Price ranges vary by tour type.
visitor_source String How the visitor arrived at the site. One of: organic (search engine or direct), paid_ad (Google Ads), hostel_referral (partner hostel links), agency (travel agency referrals).

Experiment Setup

  • Traffic split: 50/50 random assignment. Each visitor sees either version A or version B.
  • Duration: 60 days.
  • Page A (control): The original booking page design.
  • Page B (treatment): The redesigned booking page.
  • Scope: The dataset contains all visitors to either version of the English-language booking page during the experiment period.