Eric Chau's UBC MLIS Portfolio

08. System Integration – Analyze

Expedia.com

Institution

Expedia.com (Expedia Group)

Resources

The hotel listing at Expedia.com (for simplicity, only the hotel listings are considered even though Expedia offers more than hotel booking)

Description

  • Hotel listings are submitted by hotel operators using Expedia’s portal system; the hotel operators are responsible for the accuracy of their listings
    • Resource: Hotel Inventory (590k properties from more than 200 booking sites)(Expedia Group, 2022)
    • Classification
      • Guest rating, Neighbourhood, Star Rating, Price Range, etc.
    • Thesaurus
    • Controlled Vocabulary & Data Format
      • Data entry is restricted by the property management form
      • There are free form text fields (descriptions, etc.) as well as rich media such as pictures, etc.
      • Hotel availability calendar and nightly pricing

Analytical

  • The system connects hotel inventories to potential customers
  • Only available inventory will be shown; reserved inventory will be hidden from search result
  • The users will search for a property based on keyboards (hotel name, location) and date range
  • The system favours conversion (i.e., money spent on booking); it has machine learning algorithms to optimize the conversion, mostly by providing relevant search result and recommendations.
  • Key elements (e.g., pricing information, hero images, star rating, custom tag, etc.) are often favoured in visual cues for listing results; textual description and details are shown in the property detail page once the user clicks on a listing
  • Records are entered into a customer portal with onscreen validation
    • Only permissible entry can be submitted and will be objected to moderation by the Expedia team
    • Expedia’s team will refine the system based on internal researches
  • Systems at work
    • Filters are based on different Classifications
      • Mainly used as a means to refine search results
    • Search is smart enough to utilize the Thesaurus to suggest related and similar properties being searched (Greater Vancouver: NT Burnaby RT Vancouver RT YVR, etc.)
    • Controlled Vocabulary and Data Format
      • There are limits for choice of neighbourhood, number of characters in listing, and number of pictures, etc.

Rationale for the System at work

  • Expedia receives a commission for each reservation booked
  • Expedia has incentive to convert (complete booking) for each lead (potential customer)
  • Expedia collects analytics to optimize User Experience (from what to show, size of elements, location of elements, etc.) to maximize bookings.
    • Typically, leads might drop off it they couldn’t find a good match for the search criteria. As such, relevancy is often key
    • Interaction with systems (clicks on listing might imply intent; ignored listing might imply low relevancy, etc.) will be analyzed to optimize search results
    • Classification
      • Makes filtering by the key criteria (such as star rating, price range) much quicker and simpler (i.e., shows only 4 star hotels under $200 a night, etc.)
    • Thesaurus
      • Provide relevant recommendations and related terms to enhance experience, most likely through machine learning algorithms and data analysis
    • Controlled Vocabulary and Data Format
      • Ensure the listing confronts to established quality guideline based on Expedia’s research
      • Ensure the user experiences (i.e., filters and listing search result and property detail page, etc.) are mostly consistent

Reflection

I think there is some confusion re: this being a discussion of multiple standards. The BCP 47 is specifically designed to work with other already published standards such as the HTML specification and ISO-3166 country code standards, and it inherits many of the same limitations of these standards. For this particular example, we are looking only at the classification aspect of the standard.

For instance, the spoken language Cantonese, a Chinese dialect, (or the unofficial written way of the spoken dialect) cannot be currently be represented by such classification. This limitation, for example, has nothing to do with the HTML specification (which uses the BCP 47 as a way to represent a language in a HTML document but not the only usage for the BCP 47 reference) or the ISO country specification (the standard being leveraged to indicate a country, which is a small but integral part of the BCP 47 specification).