Business Case Analysis: Airbnb Customer Host Matching Study
Airbnb is a two-sided marketplace that matches guests to hosts. The booking flow at Airbnb is as follows: a guest finds an available room (listing) that they like online, and then they contact the host. Once the guest finds a listing they are interested in, there are three ways to send the host an inquiry: ¡®contact_me¡¯, ¡®book_it¡¯, or ¡®instant_book¡¯.
Upon receiving the inquiry, the host can decide whether to accept the request (for ¡®contact_me¡¯ and ¡®book_it¡¯ methods — `instant_book` is auto-accepted). One of our goals at Airbnb is to help maximize the likelihood of a successful guest-host match on our platform.
Targeted Outcomes
We at Airbnb would like to understand better what causes guests on our platform to end up with a successful booking.
Please identify, analyze, and visualize the opportunities to increase successful guest-host matching in Rio de Janeiro using the artificial datasets provided.
1. Prepare these findings to present to the market manager of Rio and your fellow data team. 1.1. In particular, questions you should address include:
o What key metrics would you propose to monitor the success of the team’s efforts in improving the guest host matching process and why? Clearly define your metric(s) and explain how each is computed in addition to visualization.
o What opportunities exist to increase the number of successful bookings in Rio de Janeiro? What segments are doing well, and what could be improved?
o What are 2-3 specific recommendations that could address these opportunities?
1.2. Demonstrate the rationale behind each recommendation AND prioritize your recommendations based on their estimated impact.
2. Think outside of the data set you were given. What other research, experiments, or approaches could help the company clarify the problem?
Suggestions
A: We look forward to business and data analysts who go beyond basic data summary to make reasoned recommendations backed by data (and clarify assumptions). Many models can be applicable in this case. However, Tensorflow (especially Artificial Neural Network) models will be highly appreciated and preferred.
B: We suggest that you browse the Airbnb website and look at listings to see the different ways that you can message a host.
C: Prepare a write-up that is possible to share in a presentation that would last ~10 -15 minutes (feel free to include more detailed slides in your appendix as needed).
Data Provided
o Contacts – contains a row for every time a user inquires about a stay at a listing in Rio de Janeiro. o id_guest_anon – id of the guest inquiring.
Code Help, Add WeChat: cstutorcs
o id_host_anon – id of the listing host to which the inquiry is made. o id_listing_anon – id of the listing to which the inquiry is made.
o ts_interaction_first – UTC timestamp of the moment the inquiry is made.
o ts_reply_at_first – UTC timestamp of the moment the host replies to the inquiry, if so.
ts_accepted_at_first – UTC timestamp of when the host accepts the inquiry, if so. o ts_booking_at – UTC timestamp of when the booking is made.
o ds_checkin_first – Date stamp of the check-in date of the inquiry.
o ds_checkout_first – Date stamp of the check-out date of the inquiry.
o m_guests – The number of guests the inquiry is for.
o m_interactions – The total number of messages sent by both the guest and host (when the data
was pulled).
o m_first_message_length_in_characters – Number of characters in the first message sent by the
guest, if a message was sent
o contact_channel_first-Thecontactchannelthroughwhichtheinquirywasmade.Oneof
{contact_me, book_it, instant_book}. *See bottom of the page for more detail*
o guest_user_stage_first – Indicates whether the user has booked before sending the inquiry (¡°past booker¡±). If the user has not booked before, then the user is new.
o Listings – contains data for every listing
o id_listing_anon – anonymized id of the listing to which the inquiry is made.
o room_type – indicates whether the room is an entire home, private room, or shared room listing_neighborhood – the neighborhood of the listing
o total_reviews – the total number of reviews of the listing (when the data was pulled).
o Users – contains data for every user
o id_user_anon – anonymized id of the user
o words_in_user_profile – the number of words in the ¡°about me¡± section of the user¡¯s Airbnb
profile (at the time of contact)
o country – origin country of the user
Further Information
There are three ways to book a listing on Airbnb:
1) contact_me – The guests write a message to the host to inquire about the listing. The host has the
option to (i) pre-approve the guest to book their place, (ii) they can reject, or (iii) write a free text message with no explicit acceptance or rejection.
If the host pre-approves, the guest can then go ahead and click to make the booking (but is not obligated to).
Code Help
2) book_it – The guest puts money down to book the place directly, but the host has to accept the reservation request. If the host accepts, the booking happens automatically. If you have used Airbnb before, this button is labeled ¡°Request to book.¡±
3) instant_book – The guest books the listing directly without needing the host to accept or reject actively (it is ddddddddddddd by the host). This shows up as a button labeled ¡°Book.¡±
Note: A host can opt-in to the `instant_book` feature. If a host does so, a guest can use the `contact_me` or `instant_book` channels for booking that particular listing but cannot use the `book_it` functionality. Alternatively, guests can use the `contact_me` or `book_it` channels if a host does not opt-in.
Computer Science Tutoring